Which of the following configuration files ____ contain S configuration information for Sendmail helping the investigator to determine where the log files reside?

The three-hour study guide for the final exam

Latest update: Fri Oct 2 10:10:51 EDT 2020

Disclaimer: This study guide attempts to touch upon the most important topics that may be covered on the final exam but does not claim to necessarily cover everything that one needs to know for the exam. Finally, don't take the three hour time window in the title literally.

Introduction

Computer security is about keeping computers, their programs, and the data they manage “safe.” Specifically, this means safeguarding three areas: confidentiality, integrity, and availability. These three are known as the CIA Triad [no relation to the Central Intelligence Agency].

ConfidentialityConfidentiality means that we do not make a system’s data and its resources [the devices it connects to and its ability to run programs] available to everyone. Only authorized people and processes should have access. Privacy specifies limits on what information can be shared with others while confidentiality provides a means to block access to such information. Privacy is a reason for confidentiality. Someone being able to access a protected file containing your medical records without proper access rights is a violation of confidentiality.Integrity

Integrity refers to the trustworthiness of a system. This means that everything is as you expect it to be: users are not imposters and processes are running correctly.

• Data integrity means that the data in a system has not been corrupted.

• Origin integrity means that the person or system sending a message or creating a file truly is that person and not an imposter.

• Recipient integrity means that the person or system receiving a message truly is that person and not an imposter.

• System integrity means that the entire computing system is working properly; that it has not been damaged or subverted. Processes are running the way they are supposed to.

Maintaining integrity means not just defending against intruders that want to modify a program or masquerade as others. It also means protecting the system against against accidental damage, such as from user or programmer errors.

AvailabilityAvailability means that the system is available for use and performs properly. A denial of service [DoS] attack may not steal data or damage any files but may cause a system to become unresponsive.

Security is difficult. Software is incredibly complex. Large systems may comprise tens or hundreds of millions of lines of code. Systems as a whole are also complex. We may have a mix of cloud and local resources, third-party libraries, and multiple administrators. If security was easy, we would not have massive security breaches year after year. Microsoft wouldn’t have monthly security updates. There are no magic solutions … but there is a lot that can be done to mitigate the risk of attacks and their resultant damage.

We saw that computer security addressed three areas of concern. The design of security systems also has three goals.

PreventionPrevention means preventing attackers from violating established security policies. It means that we can implement mechanisms into our hardware, operating systems, and application software that users cannot override – either maliciously or accidentally. Examples of prevention include enforcing access control rules for files and authenticating users with passwords.DetectionDetection detects and reports security attacks. It is particularly important when prevention mechanisms fail. It is useful because it can identify weaknesses with certain prevention mechanisms. Even if prevention mechanisms are successful, detection mechanisms are useful to let you know that attempted attacks are taking place. An example of detection is notifying an administrator that a new user has been added to the system. Another example is being notified that there have been several consecutive unsuccessful attempts to log in.RecoveryIf a system is compromised, we need to stop the attack and repair any damage to ensure that the system can continue to run correctly and the integrity of data is preserved. Recovery includes forensics, the study of identifying what happened and what was damaged so we can fix it. An example of recovery is restoration from backups.

Security engineering is the task of implementing the necessary mechanisms and defining policies across all the components of the system. Like other engineering disciplines, designing secure systems involves making compromises. A highly secure system will be disconnected from any communication network, sit in an electromagnetically shielded room that is only accessible to trusted users, and run software that has been thoroughly audited. That environment is not acceptable for most of our computing needs. We want to download apps, carry our computers with us, and interact with the world. Even in the ultra-secure example, we still need to be concerned with how we monitor access to the room, who wrote the underlying operating system and compilers, and whether authorized users can be coerced to subvert the system. Systems have to be designed with some idea of who are likely potential attackers and what the threats are. Risk analysis is used to understand the difficulty of an attack on a system, who will be affected, and what the worst thing that can happen is. A threat model is a data flow model [e.g., diagram] that identifies each place where information moves into or out of the software or between subsystems of the program. It allows you to identify areas where the most effort should be placed to secure a system.

Secure systems have two parts to them: mechanisms and policies. A policy is a description of what is or is not allowed. For example, “users must have a password to log into the system” is a policy. Mechanisms* are used to implement and enforce policies. An example of a mechanism is the software that requests user IDs and passwords, authenticates the user, and allows entry to the system only if the correct password is used.

A vulnerability is a weakness in the security system. It could be a poorly defined policy, a bribed individual, or a flaw in the underlying mechanism that enforces security. An attack is the exploitation of a vulnerability in a system. An attack vector refers to the specific technique that an attacker uses to exploit a vulnerability. Example attack vectors include phishing, keylogging, and trying common passwords to log onto a system. An attack surface is the sum of possible attack vectors in a system: all the places where an attacker might try to get into the system.

A threat is the potential adversary who may attack the system. Threats may lead to attacks.

Threats fall into four broad categories:

Disclosure: Unauthorized access to data, which covers exposure, interception, interference, and intrusion. This includes stealing data, improperly making data available to others, or snooping on the flow of data.

Deception: Accepting false data as true. This includes masquerading, which is posing as an authorized entity; substitution or insertion of includes the injection of false data or modification of existing data; repudiation, where someone falsely denies receiving or originating data.

Disruption: Some change that interrupts or prevents the correct operation of the system. This can include maliciously changing the logic of a program, a human error that disables a system, an electrical outage, or a failure in the system due to a bug. It can also refer to any obstruction that hinders the functioning of the system.

Usurpation: Unauthorized control of some part of a system. This includes theft of service as well as any misuse of the system such as tampering or actions that result in the violation of system privileges.

The Internet increases opportunities for attackers. The core protocols of the Internet were designed with decentralization, openness, and interoperability in mind rather than security. Anyone can join the Internet and send messages … and untrustworthy entities can provide routing services. It allows bad actors to hide and to attack from a distance. It also allows attackers to amass asymmetric power: harnessing more resources to attack than the victim has for defense. Even small groups of attackers are capable of mounting Distributed Denial of Service [DDoS] attacks that can overwhelm large companies or government agencies.

Adversaries can range from lone hackers to industrial spies, terrorists, and intelligence agencies. We can consider two dimensions: skill and focus. Regarding focus, attacks are either opportunistic or targeted. Opportunistic attacks are those where the attacker is not out to get you specifically but casts a wide net, trying many systems in the hope of finding a few that have a particular vulnerability that can be exploited. Targeted attacks are those where the attacker targets you specifically. The term script kiddies is used to refer to attackers who lack the skills to craft their own exploits but download malware toolkits to try to find vulnerabilities [e.g., systems with poor or default passwords, hackable cameras]. Advanced persistent threats [APT] are highly-skilled, well-funded, and determined [hence, persistent] attackers. They can craft their own exploits, pay millions of dollars for others, and may carry out complex, multi-stage attacks.

We refer to the trusted computing base [TCB] as the collection of hardware and software of a computing system that is critical to ensuring the system’s security. Typically, this is the operating system and system software. If the TCB is compromised, you no longer have assurance that any part of the system is secure. For example. the operating system may be modified to ignore the enforcement of file access permissions. If that happens, you no longer have assurance that any application is accessing files properly.

Access control

See lecture notes

Program Hijacking

Program hijacking refers to techniques that can be used to take control of a program and have it do something other than what it was intended to do. One class of techniques uses code injection, in which an adversary manages to add code to the program and change the program’s execution flow to run that code.

The best-known set of attacks are based on buffer overflow. Buffer overflow is the condition where a programmer allocates a chunk of memory [for example, an array of characters] but neglects to check the size of that buffer when moving data into it. Data will spill over into adjacent memory and overwrite whatever is in that memory.

Languages such as C, C++, and assembler are susceptible to buffer overflows since the language does not have a means of testing array bounds. Hence, the compiler cannot generate code to validate that data is only going into the allocated buffer. For example, when you copy a string using strcpy[char *dest, char *src], you pass the function only source and destination pointers. The strcpy function has no idea how big either of the buffers are.

Stack-based overflows

When a process runs, the operating system’s program loader allocates a region for the executable code and static data [called the text and data segments], a region for the stack, and a region for the heap [used for dynamic memory allocation, such as by malloc].

Just before a program calls a function, it pushes the function’s parameters onto the stack. When the call is made, the return address gets pushed on the stack. On entry to the function that was called, the function pushes the current frame pointer [a register in the CPU] on the stack, which forms a linked list to the previous frame pointer and provides an easy way to revert the stack to where it was before making the function call. The frame pointer register is then set to the current top of the stack. The function then adjusts the stack pointer to make room for hold local variables, which live on the stack. This region for the function’s local data is called the stack frame. Ensuring that the stack pointer is always pointing to the top of the stack enables the function to get interrupts or call other functions without overwriting anything useful on the stack. The compiler generates code to reference parameters and local variables as offsets from the current frame pointer register.

Before a function returns, the compiler generates code to:

• Adjust the stack back to point to where it was before the stack expanded to make room for local variables. This is done by copying the frame pointer to the stack pointer.

• Restore the previous frame pointer by popping it off the stack [so that local variables for the previous function could be referenced properly].

• Return from the function. Once the previous frame pointer has been popped off the stack, the stack pointer points to a location on the stack that holds the return address.

Simple stack overflows

Local variables are allocated on the stack and the stack grows downward in memory. Hence, the top of the stack is in lower memory than the start, or bottom, of the stack. If a buffer [e.g., char buf[128]] is defined as a local variable, it will reside on the stack. As the buffer gets filled up, its contents will be written to higher and higher memory addresses. If the buffer overflows, data will be written further down the stack [in higher memory], overwriting the contents of any other variables that were allocated for that function and eventually overwriting the saved frame pointer and the saved return address.

When this happens and the function tries to return, the return address that is read from the stack will contain garbage data, usually a memory address that is not mapped into the program’s memory. As such, the program will crash when the function returns and tries to execute code at that invalid address. This is an availability attack. If we can exploit the fact that a program does not check the bounds of a buffer and overflows the buffer, we can cause a program to crash.

Subverting control flow through a stack overflow

Buffer overflow can be used in a more malicious manner. The buffer itself can be filled with bytes of valid machine code. If the attacker knows the exact size of the buffer, she can write just the right number of bytes to write a new return address into the very same region of memory on the stack that held the return address to the parent function. This new return address points to the start of the buffer that contains the injected code. When the function returns, it will “return” to the new code in the buffer and execute the code at that location.

Off-by-one stack overflows

As we saw, buffer overflow occurs because of programming bugs: the programmer neglected to make sure that the data written to a buffer does not overflow. This often occurs because the programmer used old, unsafe functions that do not allow the programmer to specify limits. Common functions include:

- strcpy[char *dest, char *src] - strcat[char *dest, char *src] - sprintf[char *format, ...]

Each of these functions has a safe counterpart that accepts a count parameter so that the function will never copy more than count number of bytes:

- strcpy[char *dest, char *src, int count] - strcat[char *dest, char *src, int count] - sprintf[char *format, int count, ...]

You’d think this would put an end to buffer overflow problems. However, programmers may miscount or they may choose to write their own functions that do not check array bounds correctly. A common error is an off-by-one error. For example, a programmer may declare a buffer as:

char buf[128];

and then copy into it with:

for [i=0; i [time_t]expiration] { fprintf[stderr, "This software expired on %s", ctime[&expiration]]; fprintf[stderr, "This time is now %s", ctime[&now]]; } else fprintf[stderr, "You're good to go: %lu days left in your trial.\n", [expiration-now]/[60*60*24]]; return 0; }

When run, we may get output such as:

$ ./testdate This software expired on Sat Dec 31 19:00:00 2016 This time is now Sun Feb 18 15:50:44 2018

Let us write a replacement time function that always returns a fixed value that is less than the one we test for. We’ll put it in a file called time.c:

unsigned long time[] { return [unsigned long] 1483000000; }

We compile it into a shared library:

gcc -shared -fPIC time.c -o newtime.so

Now we set LD_PRELOAD and run the program:

$ export LD_PRELOAD=$PWD/newtime.so $ ./testdate You're good to go: 2 days left in your trial.

Note that our program now behaves differently and we never had to recompile it or feed it different data!

Input sanitization

The important lesson in writing code that uses any user input in forming commands is that of input sanitization. Input must be carefully validated to make sure it conforms to the requirements of the application that uses it and does not try to execute additional commands, escape to a shell, set malicious environment variables, or specify out-of-bounds directories or devices.

File descriptors

POSIX systems have a convention that programs expect to receive three open file descriptors when they start up:

• file descriptor 0: standard input

• file descriptor 1: standard output

• file descriptor 2: standard error

Functions such as printf, scanf, puts, getc and others expect these file desciptors to be available for input and output. When a program opens a new file, the operating system searches through the file descriptor table and allocates the first available unused file descriptor. Typically this will be file descriptor 3. However, if any of the three standard file descriptors are closed, the operating system will use one of those as an available, unused file descriptor.

The vulnerability lies in the fact that we may have a program running with elevated privileges [e.g., setuid root] that modifies a file that is not accessible to regular users. If that program also happens to write to the user via, say, printf, there is an opportunity to corrupt that file. The attacker simply needs to close the standard output [file descriptor 1] and run the program. When it opens its secret file, it will be given file descriptor 1 and will be able to do its read and write operations on the file. However, whenever the program will print a message to the user, the output will not be seen by the user as it will be directed to what printf assumes is the standard output: file descriptor 1. Printf output will be written onto the secret file, thereby corrupting it.

The shell command [bash, sh, or ksh] for closing the standard output file is an obscure-looking >&-. For example:

./testfile >&-

Comprehension Errors

The overwhelming majority of security problems are caused by bugs or misconfigurations. Both often stem from comprehension errors. These are mistakes created when someone – usually the programmer or administrator – does not understand the details and every nuance of what they are doing. Some example include:

• Not knowing all possible special characters that need escaping in SQL commands.

• Not realizing that the standard input, output, or error file descriptors may be closed.

• Not understanding how access control lists work or how to configure mandatory access control mechanisms such as type enforcement correctly.

If we consider the Windows CreateProcess function, we see it is defined as:

BOOL WINAPI CreateProcess[ _In_opt_ LPCTSTR lpApplicationName, _Inout_opt_ LPTSTR lpCommandLine, _In_opt_ LPSECURITY_ATTRIBUTES lpProcessAttributes, _In_opt_ LPSECURITY_ATTRIBUTES lpThreadAttributes, _In_ BOOL bInheritHandles, _In_ DWORD dwCreationFlags, _In_opt_ LPVOID lpEnvironment, _In_opt_ LPCTSTR lpCurrentDirectory, _In_ LPSTARTUPINFO lpStartupInfo, _Out_ LPPROCESS_INFORMATION lpProcessInformation];

We have to wonder whether a programmer who does not use this frequently will take the time to understand the ramifications of correctly setting process and thread security attributes, the current directory, environment, inheritance handles, and so on. There’s a good chance that the programmer will just look up an example on places such as github.com or stackoverflow.com and copy something that seems to work, unaware that there may be obscure side effects that compromise security.

As we will see in the following sections, comprehension errors also apply to the proper understanding of things as basic as various ways to express characters.

Directory parsing

Some applications, notably web servers, accept hierarchical filenames from a user but need to ensure that they restrict access only to files within a specific point in the directory tree. For example, a web server may need to ensure that no page requests go outside of /home/httpd/html.

An attacker may try to gain access by using paths that include .. [dot-dot], which is a link to the parent directory. For example, an attacker may try to download a password file by requesting

//poopybrain.com/../../../etc/passwd

The hope is that the programmer did not implement parsing correctly and might try simply suffixing the user-requested path to a base directory:

"/home/httpd/html/" + "../../../etc/passwd"

to form

/home/httpd/html/../../../etc/passwd

which will retrieve the password file, /etc/passwd.

A programmer may anticipate this and check for dot-dot but has to realize that dot-dot directories can be anywhere in the path. This is also a valid pathname but one that should be rejected for trying to escape to the parent:

//poopybrain.com/419/notes/../../416/../../../../etc/passwd

Moreover, the programmer cannot just search for .. because that can be a valid part of a filename. All three of these should be accepted:

//poopybrain.com/419/notes/some..other..stuff/ //poopybrain.com/419/notes/whatever../ //poopybrain.com/419/notes/..more.stuff/

Also, extra slashes are perfectly fine in a filename, so this is acceptable:

//poopybrain.com/419////notes///////..more.stuff/

The programmer should also track where the request is in the hierarchy. If dot-dot doesn’t escape above the base directory, it should most likely be accepted:

//poopybrain.com/419/notes/../exams/

These are not insurmountable problems but they illustrate that a quick-and-dirty attempt at filename processing may be riddled with bugs.

Unicode parsing

If we continue on the example of parsing pathnames in a web server, let us consider a bug in early releases of Microsoft’s IIS [Internet Information Services, their web server]. IIS had proper pathname checking to ensure that attempts to get to a parent are blocked:

//www.poopybrain.com/scripts/../../winnt/system32/cmd.exe

Once the pathname was validated, it was passed to a decode function that decoded any embedded Unicode characters and then processed the request.

The problem with this technique was that non-international characters [traditional ASCII] could also be written as Unicode characters. A “/” could also be written in HTML as its hexadecimal value, %2f [decimal 47]. It could also be represented as the two-byte Unicode sequence %c0%af.

The reason for this stems from the way Unicode was designed to support compatibility with one-byte ASCII characters. This encoding is called UTF–8. If the first bit of a character is a 0, then we have a one-byte ASCII character [in the range 0..127]. However, if the first bit is a 1, we have a multi-byte character. The number of leading 1s determine the number of bytes that the character takes up. If a character starts with 110, we have a two-byte Unicode character.

With a two-byte character, the UTF–8 standard defines a bit pattern of

110a bcde 10fg hijk

The values a-k above represent 11 bits that give us a value in the range 0..2047. The “/” character, 0x2f, is 47 in decimal and 0010 1111 in binary. The value represents offset 47 into the character table [called codepoint in Unicode parlance]. Hence we can represent the “/” as 0x2f or as the two byte Unicode sequence:

1100 0000 1010 1111

which is the hexadecimal sequence %c0%af. Technically, this is disallowed. The standard states that codepoints less than 128 must be represented as one byte but the two byte sequence is supported by most Unicode parsers. We can also construct a valid three-byte sequence too.

Microsoft’s bug was that they ignored parsing %c0%af as being equivalent to a / because it should not have been used to represent the character. However, the Unicode parser was happy to translate it and attackers were able to use this to access any file in on a server running IIS. This bug also gave attackers the ability to invoke cmd.com, the command interpreter, and execute any commands on the server.

After Microsoft fixed the multi-byte Unicode bug, another problem came up. The parsing of escaped characters was recursive, so if the resultant string looked like a Unicode hexadecimal sequence, it would be re-parsed.

As an example of this, let’s consider the backslash [\], which Microsoft treats as equivalent to a slash [/] in URLs since their native pathname separator is a backlash[3].

The backslash can be written in a URL in hexadecimal format as %5c. The “%” character can be expressed as %25. The “5” character can be expressed as %35. The “c” character can be expressed as %63. Hence, if the URL parser sees the string %%35c, it would expand the %35 to the character “5”, which would result in %5c, which would then be converted to a \. If the parser sees %25%35%63, it would expand each of the %nn components to get the string %5c, which would then be converted to a \. As a final example, if the parser comes across %255c, it will expand %25 to % to get the string %5c, which would then be converted to a \.

It is not trivial to know what a name relates to but it is clear that all conversions have to be done before the validity of the pathname is checked. As for checking the validity of the pathname in an application, it is error-prone. The operating system itself parses a pathname a component at a time, traversing the directory tree and checking access rights as it goes along. The application is trying to recreate a similar action without actually traversing the file system but rather by just parsing the name and mapping it to a subtree of the file system namespace.

TOCTTOU attacks

TOCTTOU stands for Time of Check to Time of Use. If we have code of the form:

if I am allowed to do something then do it

we may be exposing ourselves to a race condition. There is a window of time between the test and the action. If an attacker can change the condition after the check then the action may take place even if the check should have failed.

One example of this is the print spooling program, lpr. It runs as a setuid program with root privileges so that it can copy a file from a user’s directory into a privileged spool directory that serves as a queue of files for printing. Because it runs as root, it can open any file, regardless of permissions. To keep the user honest, it will check access permissions on the file that the user wants to print and then, only if the user has legitimate read access to the file, it will copy it over to the spool directory for printing. An attacker can create a link to a readable file and then run lpr in the background. At the same time, he can change the link to point to a file for which he does not have read access. If the timing is just perfect, the lpr program will check access rights before the file is re-linked but will then copy the file for which the user has no read access.

Another example of the TOCTTOU race condition is the set of temporary filename creation functions [tempnam, tempnam, mktemp, GetTempFileName, etc.]. These functions create a unique filename when they are called but there is no guarantee that an attacker doesn’t create a file with the same name before that filename is used. If the attacker creates and opens a file with the same name, she will have access to that file for as long as it is open, even if the user’s program changes access permissions for the file later on.

The best defense for the temporary file race condition is to use the mkstemp function, which creates a file based on a template name and opens it as well, avoiding the race condition between checking the uniqueness of the name and opening the file.

App confinement

Two lessons we learned from experience are that applications can be compromised and that applications may not always be trusted. Server applications, in particular, such as web servers and mail servers have been compromised over and over again. This is particularly harmful as they often run with elevated privileges and on systems on which normal users do not have accounts. The second category of risk is that we may not always trust an application. We trust our web server to work properly but we cannot necessarily trust that the game we downloaded from some unknown developer will not try to upload our files, destroy our data, or try to change our system configuration. In fact, unless we have the ability to scrutinize the codebase of a service, we will not know for sure if it tries to modify any system settings or writes files to unexpected places.

With this resignation to security in mind, we need to turn our attention to limiting the resources available to an application and making sure that a misbehaving application cannot harm the rest of the system. These are the goals of confinement.

Our initial thoughts to achieving confinement may involve proper use of access controls. For example, we can run server applications as low-privilege users and make sure that we have set proper read/write/execute permissions on files, read/write/search permissions on directories, or even set up role-based policies.

However, access controls usually do not give us the ability to set permissions for “don’t allow access to anything else.” For example, we may want our web server to have access to all files in /home/httpd but nothing outside of that directory. Access controls do not let us express that rule. Instead, we are responsible for changing the protections of every file on the system and making sure it cannot be accessed by “other”. We also have to hope that no users change those permissions. In essence, we must disallow the ability for anyone to make files publicly accessible because we never want our web server to access them. We may be able to use mandatory access control mechanisms if they are available but, depending on the system, we may not be able to restrict access properly either. More likely, we will be at risk of comprehension errors and be likely to make a configuration error, leaving parts of the system vulnerable. To summarize, even if we can get access controls to help, we will not have high assurance that they do.

Access controls also only focus on protecting access to files and devices. A system has other resources, such as CPU time, memory, disk space, and network. We may want to control how much of all of these an application is allowed to use. POSIX systems provide a setrlimit system call that allows one to set limits on certain resources for the current process and its children. These controls include the ability to set file size limits, CPU time limits, various memory size limits, and maximum number of open files.

We also may want to control the network identity for an application. All applications share the same IP address on a system but this may allow a compromised application to exploit address-based access controls. For example, you may be able to connect to or even log into system that believe you are a trusted computer. An exploited application may end up confusing network intrusion detection systems.

Just limiting access through resource limits and file permissions is also insufficient for services that run as root. If an attacker can compromise an app and get root access to execute arbitrary functions, she can change resource limits [just call setrlimit with different values], change any file permissions, and even change the IP address and domain name of the system.

In order to truly confine an application, we would like to create a set of mechanisms that enforce access controls to all of a system’s resources, are easy to use so that we have high assurance in knowing that the proper restrictions are in place, and work with a large class of applications. We can’t quite get all of this yet but we can come close.

chroot

The oldest app confinement mechanism is Unix’s chroot system call and command, originally introduced in 1979 in the seventh edition[1]. The chroot system call changes the root directory of the calling process to the directory specified as a parameter.

chroot["/home/httpd/html"];

Sets the root of the file system to /home/httpd/html for the process and any processes it creates. The process cannot see any files outside that subset of the directory tree. This isolation is often called a chroot jail.

Jailkits

If you run chroot, you will likely get an error along the lines of:

# chroot newroot chroot: failed to run command ‘/bin/bash’: No such file or directory

This is because /bin/bash is not within the root [in this case, the newroot directory]. You’ll then create a bin subdirectory and try running chroot again and get the same error:

# mkdir newroot/bin # ln /bin/bash newroot/bin/bash # chroot newroot chroot: failed to run command ‘/bin/bash’: No such file or directory

You’ll find that is also insufficient and that you’ll need to bring in the shared libraries that /bin/bash needs by mounting /lib, /lib64, and /usr/lib within that root just to enable the shell to run. Otherwise, it cannot load the libraries it needs since it cannot see above its root [i.e., outside its jail]. To simplify this process, a jailkit simplifies the process of setting up a chroot jail by providing a set of utilities to make it easier to create the desired environment within the jail and populate it with basic accounts, commands, and directories.

Problems with chroot

Chroot only limits access to the file system namespace. It does not restrict access to resources and does not protect the machine’s network identity. Applications that are compromised to give the attacker root access make the entire system vulnerable since the attacker has access to all system calls.

Chroot is available only to administrators. If this was not the case then any user would be able to get root access within the chroot jail. You would: 1. Create a chroot jail 2. Populate it with the shell program and necessary support libraries 3. Link the su command [set user, which allows you to authenticate to become any user] 4. Create password files within the jail with a known password for root. 5. Use the chroot command to enter the jail. 6. Run su root to become the root user. The command will prompt you for a password and validate it against the password file. Since all processes run within the jail, the password file is the one you set up.

You’re still in the jail but you have root access.

Escaping from chroot

If someone manages to compromise an application running inside a chroot jail and become root, they are still in the jail but have access to all system calls. For example, they can send signals to kill all other processes or shut down the system. This would be an attack on availability.

Attaining root access also provides a few ways of escaping the jail. On POSIX systems, all non-networked devices are accessible as files within the filesystem. Even memory is accessible via a file [/dev/mem]. An intruder in a jail can create a memory device [character device, major number = 1, minor number = 1]:

mknod mem c 1 1

With the memory device, the attacker can patch system memory to change the root directory of the jail. More simply, an attacker can create a block device with the same device numbers as that of the main file system. For example, the root file system on my Linux system is /dev/sda1 with a major number of 8 and a minor number of 1. An attacker can recreate that in the jail:

mknod rootdisk b 8 1

and then mount it as a file system within the jail:

mount -t ext4 rootdisk myroot

Now the attacker, still in the jail, has full access to the entire file system, which is as good as being out of the jail. He can add user accounts, change passwords, delete log files, run any commands, and even reboot the system to get a clean login.

FreeBSD Jails

Chroot was good in confining the namespace of an application but useless against providing security if an application had root access and did nothing to restrict access to other resources.

FreeBSD Jails are an enhancement to the idea of chroot. Jails provide a restricted filesystem namespace, just like chroot does, but also place restrictions on what processes are allowed to do within the jail, including selectively removing privileges from the root user in the jail. For example, processes within a jail may be configured to:

• Bind only to sockets with a specified IP address and specific ports
• Communicate only with other processes within the jail and none outside
• Not be able to load kernel modules, even if root
• Have restricted access to system calls that include:
  • Ability to create raw network sockets
  • Ability to create devices
  • Modify the network configuration
  • Mount or unmount filesystems

FreeBSD Jails are a huge improvement over chroot since known escapes, such as creating devices and mounting filesystems and even rebooting the system are disallowed. Depending on the application, policies may be coarse. The changed root provides all or nothing access to a part of the file system. This does not make Jails suitable for applications such as a web browser, which may be untrusted but may need access to files outside of the jail. Think about web-based applications such as email, where a user may want to upload or download attachments. Jails also do not prevent malicious apps from accessing the network and trying to attack other machines … or from trying to crash the host operating system. Moreover, FreeBSD Jails is a BSD-only solution. With an estimated 0.95…1.7% share of server deployments, it is a great solution on an operating system that is not that widely used.

Linux namespaces, cgroups, and capabilities

Linux’s answer to FreeBSD Jails was a combination of three elements: control groups, namespaces, and capabilities.

Control groups [cgroups]

Linux control groups, also called cgroups, allow you to allocate resources such as CPU time, system memory, disk bandwidth, network bandwidth, and the ability to monitor resource usage among user-defined groups of processes. This allows, for example, an administrator to allocate a larger share of the processor to a critical server application.

An administrator creates one or more cgroups and assigns resource limits to each of them. Then any application can be assigned to a control group and will not be able to use more than the resource limits configured in that control group. Applications are unaware of these limits. Control groups are organized in a hierarchy similar to processes. Child cgroups inherit some attributes from the parents.

Linux namespaces

Chroot only restricted the filesystem namespace. The filesystem namespace is the best known namespace in the system but not the only one. Linux namespaces Namespaces provide control over how processes are isolated in the following namespaces:

NamespaceDescriptionControls

IPC	System V IPC, POSIX message queues	Objects created in an IPC namespace are only visible to other processes in that namespace [CLONE_NEWIPC]
Network	Network devices, stacks, ports	Isolates IP protocol stacks, IP routing tables, firewalls, socket port numbers [ CLONE_NEWNET]
Mount	Mount points	A set of processes can have their own distinct mount points and view of the file system [CLONE_NEWNS]
PID	Process IDs	Processes in different PID namespaces can have their process IDs – the child cannot see parent processes or other namespaces [CLONE_NEWPID]
User	User & group IDs	Per-namespace user/group IDs. Also, you can be root in a namespace but have restricted privileges [ CLONE_NEWUSER ]
UTS	host name and domain name	setting hostname and domainname will not affect rest of the system [CLONE_NEWUTS]
Cgroup	control group	Sets a new control group for a process [CLONE_NEWCGROUP]

A process can dissociate any or all of these namespaces from its parent via the unshare system call. For example, by unsharing the PID namespace, a process gets a no longer sees other processes and will only see itself and any child processes it creates.

The Linux clone system call is similar to fork in that it creates a new process. However, it allows you to pass flags that will specify which parts of the execution context will be shared with the parent. For example, a cloned process may choose to share memory and open file descriptors, which will make it behave like threads. It can also choose to share – or not – any of the elements of the namespace.

Capabilities

A problem that FreeBSD Jails tackled was that of restricting the power of root inside a Jail. You could be a root user but still disallowed from executing certain system calls. POSIX [Linux] capabilities[2] tackle this issue as well.

Traditionally, Unix systems distinguished privileged versus unprivileged processes. Privileged processes were those that ran with a user ID of 0, called the root user. When running as root, the operating system would allow access to all system calls and all access permission checks were bypassed. You could do anything.

Linux capabilities identify groups of operations, called capabilities, that can be controlled independently on a per-thread basis. The list is somewhat long, 38 groups of controls, and includes capabilities such as:

• CAP_CHOWN: make arbitrary changes to file UIDs and GIDs
• CAP_DAC_OVERRIDE: bypass read/write/execute checks
• CAP_KILL: bypass permission checks for sending signals
• CAP_NET_ADMIN: network management operations
• CAP_NET_RAW: allow RAW sockets
• CAP_SETUID: arbitrary manipulation of process UIDs
• CAP_SYS_CHROOT: enable chroot

The kernel keeps track of four capability sets for each thread. A capability set is a list of zero or more capabilities. The sets are:

• Permitted: If a capability is not in this set, the thread or its children can never require that capability. This limits the power of what a process and its children can do.

• Inheritable: These capabilities will be inherited when a thread calls execve to execute a program [POSIX programs are executed with the same thread; we are not creating a new process]

• Effective: This is the current set of capabilities that the thread is using. The kernel uses these to perform permission checks.

• Ambient: This is similar to Inheritable and contains a set of capabilities that are preserved across an execve of a program that is not privileged. If a setuid or setgid program is run, will clear the ambient set. These are created to allow a partial use of root features in a controlled manner. It is useful for user-level device drivers or software that needs a specific privilege [e.g., for certain networking operations].

A child process created via fork [the standard way of creating processes] will inherit copies of its parent’s capability sets following the rules of which capabilities have been marked as inheritable.

A set of capabilities can be assigned to an executable file by the administrator. They are stored as a file’s extended attributes [along with access control lists, checksums, and arbitrary user-defined name-value pairs]. When the program runs, the executing process may further restrict the set of capabilities under which it operates if it chooses to do so [for example, after performing an operation that required the capability and knowing that it will no longer need to do so].

From a security point of view, the key concept of capabilities is that they allow us to provide limited elevation of privileges to a process. A process does not need to run as root [user ID 0] but can still be granted very specific privileges. For example, we can grant the ping command the ability to access raw sockets so it can send an ICMP ping message on the network but not have any other administrative powers. The application does not need to run as root and even if an attacker manages to inject code, the opportunities for attack will be restricted.

The Linux combination of cgroups, namespaces, and capabilities provides a powerful set of mechanisms to

1. Set limits on the system resources [processor, disk, network] that a group of processes will use.

2. Constrain the namespace, making parts of the filesystem or the existence of other processes or users invisible.

3. Give restricted privileges to specific applicatiosn so they do not need to run as root.

This enables us to create stronger jails and have a fine degree of control as to what processes are or are not allowed to do in that jail.

While bugs have been found these mechanisms, the more serious problem is that of comprehension. The system has become far, far more complex than it was in the days of chroot. A user has to learn quite a lot to use these mechanisms properly. Failure to understand their behavior fully can create vulnerabilities. For example, namespaces do not prohibit a process from making privileged system calls. They simply limit what a process can see. A process may not be able to send a kill signal to another process only because it does not share the same process ID namespace.

Together with capabilities, namespaces allow a restricted environment that also places limits on the abilities to perform operations even if a process is granted root privileges. This enables ordinary users to create namespaces. You can create a namespace and even create a process running as a root user [UID 0] within that namespace but it will have no capabilities beyond those that were granted to the user; the user ID of 0 gets mapped by the kernel to a non-privileged user.

Containers

Software rarely lives as an isolated application. Some software requires multiple applications and most software relies on the installation of other libraries, utilities, and packages. Keeping track of these dependencies can be difficult. Worse yet, updating one shared component can sometimes cause another application to break. What was needed was a way to isolate the installation, execution, and management of multiple software packages that run on the same system.

Various attempts were undertaken to address these problems.

1. The most basic was to fix problems when they occurred. This required carefully following instructions for installation, update, and configuration of software and extensive testing of all services on the system when anything changed. Should something break, the service would be unavailable until the problems were fixed.

2. A drastic, but thorough, approach to isolation was to simply run each service on its own computer. That avoids conflicts in library versions and other dependencies. However, it is an expensive solution, is cumbersome, and is often overkill in most environments.

3. Finally, administrators could deploy virtual machines. This is a technology that allows one to run multiple operating systems on one computer and gives the illusion of services running on distinct systems. However, this is a heavyweight solution. Every service needs its own installation of the operating system and all supporting software for the service as well as standard services [networking, device management, shell, etc.]. It is not efficient in terms of CPU, disk, or memory resources – or even administration effort.

Containers are a mechanism that were originally created not for security but to make it easy to package, distribute, relocate, and deploy collections of software. The focus of containers is not to enable end users to install and run their favorite apps but rather for administrators to be able to deploy a variety of services on a system. A container encapsulates all the necessary software for a service, all of its dependencies, and its configuration into one package that can be easily passed around, installed, and removed.

In many ways, a container feels like a virtual machine. Containers provide a service with a private process namespace, its own network interface, and its own set of libraries to avoid problems with incompatible versions used by other software. Containers also allow an administrator to give the service restricted powers even if it runs with root [administrator] privileges. Unlike a virtual machine, however, multiple containers on one system all share the same operating system and kernel modules.

Containers are not a new mechanism. They are implemented using Linux’s control groups, namespaces, and capabilities to provide resource control, isolation, and privilege control, respectively. They also make use of a copy on write file system. This makes it easy to create new containers where the file system can track the changes made by that container over a clean base version of a file system. Containers can also take advantage of AppArmor, which is a Linux kernel module that provides a basic form of mandatory access controls based on the pathnames of files. It allows an administrator to restrict the ability of a program to access specific files even within its file system namespace.

The best-known and first truly popular container framework is Docker. A Docker Image is a file format that creates a package of applications, their supporting libraries, and other needed files. This image can be stored and deployed on many environments. Docker made it easy to deploy containers using git-like commands [docker push, docker commit] and also to perform incremental updates. By using a copy on write file system, Docker images can be kept immutable [read-only] while any changes to the container during its execution are stored separately.

As people found Docker to be useful, the next design goal was to make it easier to manage containers across a network of many computers. This is called container orchestration. There are many solutions for this, including Apache Mesos, Kubernetes, Nomad, and Docker Swarm. The best known of these is kubernetes, which was designed by Google. It coordinates storage of containers, failure of hardware and containers, and dynamic scaling: deploying the container on more machines to handle increased load. Kubernetes is coordination software, not a container system; it uses the Docker framework to run the actual container.

Even though containers were designed to simplify software deployment rather than provide security to services, they do offer several benefits in the area of security:

• They make use of namespaces, cgroups, and capabilities with restricted capabilities configured by default. This provides isolation among containers.

• Containers provide a strong separation of policy [defined by the container configuration] from the enforcement mechanism [handled by the operating system].

• They improve availability by providing the ability to have a watchdog timer monitor the running of applications and restarting them if necessary. With orchestration systems such as Kubernetes, containers can be re-deployed on another system if a computer fails.

• The environment created by a container is reproducible. The same container can be deployed on multiple systems and tested in different environments. This provides consistency and aids in testing and ensuring that the production deployment matches the one used for development and test. Moreover, it is easy to inspect exactly how a container is configured. This avoids problems encountered by manual installation of components where an administrator may forget to configure something or may install different versions of a required library.

• While containers add nothing new to security, they help avoid comprehension errors. Even default configurations will provide improved security over the defaults in the operating system and configuring containers is easier than learning and defining the rules for capabilities, control groups, and namespaces. Administrators are more likely to get this right or import containers that are already configured with reasonable restrictions.

Containers are not a security panacea. Because all containers run under the same operating system, any kernel exploits can affect the security of all containers. Similarly, any denial of service attacks, whether affecting the network or monopolizing the processor, will impact all containers on the system. If implemented and configured properly, capabilities, namespaces, and control groups should ensure that privilege escalation cannot take place. However, bugs in the implementation or configuration may create a vulnerability. Finally, one has to be concerned with the integrity of the container itself. Who configured it, who validated the software inside of it, and is there a chance that it may have been modified by an adversary either at the server or in transit?

Virtual Machines

As a general concept, virtualization is the addition of a layer of abstraction to physical devices. With virtual memory, for example, a process has the impression that it owns the entire memory address space. Different processes can all access the same virtual memory location and the memory management unit [MMU] on the processor maps each access to the unique physical memory locations that are assigned to the process.

Process virtual machines present a virtual CPU that allows programs to execute on a processor that does not physically exist. The instructions are interpreted by a program that simulates the architecture of the pseudo machine. Early pseudo-machines included o-code for BCPL and P-code for Pascal. The most popular pseudo-machine today is the Java Virtual Machine [JVM]. This simulated hardware does not even pretend to access the underlying system at a hardware level. Process virtual machines will often allow “special” calls to invoke system functions or provide a simulation of some generic hardware platform.

Operating system virtualization is provided by containers, where a group of processes is presented with the illusion of running on a separate operating system but in reality shares the operating system with other groups of processes – they are just not visible to the processes in the container.

System virtual machines allow a physical computer to act like several real machines with each machine running its own operating system [on a virtual machine] and applications that interact with that operating system. The key to this machine virtualization is to not allow each operating system to have direct access to certain privileged instructions in the processor. These instructions would allow an operating system to directly access I/O ports, MMU settings, the task register, the halt instruction and other parts of the processor that could interfere with the processor’s behavior and with the other operating systems on the system. Instead, a trap and emulate approach is used. Privileged instructions, as well as system interrupts, are caught by the Virtual Machine Monitor [VMM], also known as a hypervisor. The hypervisor arbitrates access to physical resources and presents a set of virtual device interfaces to each guest operating system [including the memory management unit, I/O ports, disks, and network interfaces]. The hypervisor also handles preemption. Just as an operating system may suspend a process to allow another process to run, the hypervisor will suspend an operating system to give other operating systems a chance to run.

The two configurations of virtual machines are hosted virtual machines and native virtual machines. With a hosted virtual machine [also called a type 2 hypervisor], the computer has a primary operating system installed that has access to the raw machine [all devices, memory, and file system]. This host operating system does not run in a virtual environment. One or more guest operating systems can then be run on virtual machines. The VMM serves as a proxy, converting requests from the virtual machine into operations that get sent to and executed on the host operating system. A native virtual machine [also called a type 1 hypervisor] is one where there is no “primary” operating system that owns the system hardware. The hypervisor is in charge of access to the devices and provides each operating system drivers for an abstract view of all the devices.

Security implications

Unlike app confinement mechanisms such as jails, containers, or sandboxes, virtual machines enable isolation all the way through the operating system. A compromised application, even with escalated privileges, can wreak havoc only within the virtual machine. Even compromises to the operating system kernel are limited to that virtual machine. However, a compromised virtual machine is not much different form having a compromised physical machine sitting inside your organization: not desirable and capable of attacking other systems in your environment.

Multiple virtual machines are usually deployed on one physical system. In cases such as cloud services [e.g., such as those provided by Amazon], a single physical system may host virtual machines from different organizations or running applications with different security requirements. If a malicious application on a highly secure system can detect that it is co-resident on a computer that is hosting another operating system and that operating system provides fewer restrictions, the malware may be able to create a covert channel to communicate between the highly secure system with classified data and the more open system. A covert channel is a general term to describe the the ability for processes to communicate via some hidden mechanism when they are forbidden by policy to do so. In this case, the channel can be created via a side channel attack. A side channel is the ability to get or transmit information using some aspects of a system’s behavior, such as changes in power consumption, radio emissions, acoustics, or performance. For example, processes on both systems, even though they are not allowed to send network messages, may create a means of communicating by altering and monitoring system load. The malware on the classified VM can create CPU-intensive task at specific times. Listener software on the unclassified VM can do CPU-intensive tasks at a constant rate and periodically measure their completion times. These completion times may vary based on whether the classified system is doing CPU-intensive work. The variation in completion times creates a means of sending 1s and 0s and hence transmitting a message.

Application Sandboxing

The goal of an application sandbox is to provide a controlled and restricted environment for code execution. This can be useful for applications that may come from untrustworthy sources, such as games from unknown developers or software downloaded from dubious sites. The program can run with minimal risk of causing widespread damage to the system. Sandboxes are also used by security researchers to observe how software behaves: what the program trying to do and whether it is attempting to access any resources in a manner that is suspicious for the application. This can help identify the presence of malware within a program. The sandbox defines and enforces what an individual application is allowed to do while executing in within its sandbox.

We previously looked at isolation via jails and containers, which use mechanisms that include namespaces, control groups, and capabilities. These constitute a widely-used form of sandboxing. However, these techniques focus on isolating an application [or group of processes] from other processes, restricting access to parts of the file system, and/or providing a separate network stack with a new IP address.

While this is great for running services without the overhead of deploying virtual machines, it does not sufficiently address the basic needs of running normal applications. We want to protect users from their applications: give users the ability to run apps but restrict what those apps can do on a per-app basis.

For example, you may want to make sure that a program accesses only files under your home directory with a suffix of “.txt”, and only for reading, without restricting the entire file system namespace as chroot would do, which would require creating a separate directory structure for shared libraries and other standard components the application may need. As another example, you might want an application to have access only to TCP networking. With a mechanism such as namespaces, we cannot exercise control over the names of files that an application can open or their access modes. Namespaces also do not allow us to control how the application interacts with the network. Capabilities allow us to restrict what a process running with root privileges can do but offer no ability to restrict more fundamental operations, such as denying a process the ability to read a file even if that file has read access enabled. The missing ingredient is rule-based policies to define precisely what system calls an application can invoke – down to the parameters of the system calls of interest.

Instead of building a jail [a container], we will add an extra layer of access control. An application will have the same view of the operating system as any other application but will be restricted in what it can do.

Sandboxing is currently supported on a wide variety of platforms at either the kernel or application level. We will examine four types of application sandboxes:

1. User-level validation
2. OS support
3. Browser-based application sandboxing
4. The Java sandbox

Note that there are many other sandbox implementations. This is just a representative sampling.

Application sandboxing via system call interposition & user-level validation

Applications interact with their environment via system calls to the operating system. Any interaction that an application needs to do aside from computation, whether legitimate or because it has been compromised, must be done through system calls: accessing files or devices, changing permissions, accessing the network, talking with other processes, etc.

An application sandbox will allow us to create policies that define which system calls are permissible to the application and in what way they can be used.

If the operating system does not provide us with the required support and we do not have the ability to recompile an application to force it to use alternate system call libraries, we can rely on system call interposition to construct a sandbox. System call interposition is the process of intercepting an app’s system calls and performing additional operations. The technique is also called hooking. In the case of a sandbox, it will intercept a system call, inspect its parameters, and decide whether to allow the system call to take place or return an error.

Example: Janus

One example of doing validation at the user level is the Janus sandboxing system, developed at UC Berkeley, originally for SunOS but later ported to Linux. Janus uses a loadable, lightweight, kernel module called mod_janus. The module initializes itself by setting up hooks to redirect system call requests to to itself. A hook is simply a mechanism that redirects an API request somewhere else and allows it to return back for normal processing. For example, a function can be hooked to simply log the fact that it has been called. The Janus kernel module copies the system call table to redirect the vector of calls to the mod_janus.

A user-configured policy file defines the allowable files and network operations for each sandboxed application. Users run applications through a Janus launcher/monitor program, which places the application in the sandbox. The monitor parses the policy file and spawns a child process for the user-specified program. The child process executes the actual application. The parent Janus process serves as the monitor, running a policy engine that receives system call notifications and decides whether to allow or disallow the system call.

Whenever a sandboxed application makes a system call, the call is redirected by the hook in the kernel to the Janus kernel module. The module blocks the thread [it is still waiting for the return from the system call] and signals the user-level Janus process that a system call has been requested. The user-level Janus process’ policy engine then requests all the necessary information about the call [calling process, type of system call, parameters]. The policy engine makes a policy decision to determine whether, based on the policy, the process should be permitted to make the system call. If so, the system call is directed back to the operating system. If not, an error code is returned to the application.

Challenges of user-level validation

The biggest challenge with implementing Janus is that the user-level monitor must mirror the state of the operating system. If the child process forks a new process, the Janus monitor must also fork. It needs to keep track of not just network operations but the proper sequencing of the steps in the protocol to ensure that no improper actions are attempted on the network. This is a sequence of socket, bind, connect, read/write, and shutdown system calls. If one fails, chances are that the others should not be allowed to take place. However, the Janus monitor does not have the knowledge of whether a particular system call succeeded or not; approved calls are simply forwarded from the module to the kernel for processing. Failure to handle this correctly may enable attack vectors such as trying to send data on an unconnected socket.

The same applies with file operations. If a file failed to open, read and write operations should not be allowed to take place. Keeping track of state also gets tricky if file descriptors are duplicated [e.g., via the dup2 system call]; it is not clear whether any requested file descriptor is a valid one or not.

Pathname parsing of file names has to be handled entirely by the monitor. We earlier examined the complexities of processing "../" sequences in pathnames. Janus has to do this in order to validate any policies on permissible file names or directories. It also has to keep track of relative filenames since the application may change the current directory at any time via the chdir system call. This means Janus needs to intercept chdir requests and process new pathnames within the proper context. Moreover, the application may change its entire namespace if the process calls chroot.

File descriptor can cause additional problems. A process can pass an open file descriptor to another process via UNIX domain sockets, which can then use that file descriptor [via a sendfd and recv_fd set of calls]. Janus would be hard-pressed to know that this happened since that would require understanding the intent of the underlying sendmsg system calls and cmsg directives.

In addition to these difficulties, user-level validation suffers from possible TOCTTOU [time-of-check-to-time-of-use] race conditions. The environment present when Janus validates a request may change by the time the request is processed.

Application sandboxing with integrated OS support

The better alternative to having a user-level process decide on whether to permit system calls is to incorporate policy validation in the kernel. Some operating systems provide kernel support for sandboxing. These include the Android Application Sandbox, the iOS App Sandbox, the macOS sandbox, and AppArmor on Linux. Microsoft introduced the Windows Sandbox in December 2018, but this functions far more like a container than a traditional application sandbox, giving the process an isolated execution environment.

Seccomp-BPF

Seccomp-BPF, which stands for SECure COMPuting with Berkeley Packet Filters, is a sandboxing framework that is available on Linux systems. It allows the user to attach a system call filter to a process and all of the descendants of that process. Users can enumerate allowable system calls and also allow or disallow access to specific files or network protocols. Seccomp has been a core part of the Android security since the release of Android O in August 2017.

Seccomp uses the Berkeley Packet Filter [BPF] interpreter, which is a framework that was initially created for network socket filtering. With socket filtering, a user can create a filter to allow or disallow certain types of data to come through the socket. Since BPF is a framework that was initially created for sockets, seccomp sends “packets” that represent system calls to the BPF [Berkeley Packet Filter] interpreter. The filter allows the user to define rules that are applied to these system calls. These rules enable the inspection of each system call and its arguments and take subsequent action. Actions include allowing the call to run or not. If the call is not permitted, rules can specify whether an error is returned to the process, a SIGSYS signal is sent, or whether the process gets killed.

Seccomp is not designed to serve as a complete sandbox solution but is a tool for building sandboxes. For further process isolation, it can be used with other components, such as namespaces, capabilities, and control groups. The biggest downside of seccomp is the use of the BPF. BPF is a full interpreter – a processor virtual machine – that supports reading/writing registers, scratch memory operations, arithmetic, and conditional branches. Policies are compiled into BPF instructions before they are loaded into the kernel. It provides a low-level interface and the rules are not simple condition-action definitions. System calls are referenced by numbers, so it is important to check the system architecture in the filter as Linux system call numbers may vary across architectures. Once the user gets past this, the challenge is to apply the principle of least privilege effectively: restrict unnecessary operations but ensure that the program still functions correctly, which includes things like logging errors and other extraneous activities.

The Apple Sandbox

Conceptually, Apple’s sandbox is similar to seccomp in that it is a kernel-level sandbox, although it does not use the Berkeley Packet Filter. The sandbox comprises:

• User-level library functions for initializing and configuring the sandbox for a process
• A server process for handling logging from the kernel
• A kernel extension that uses the TrustedBSD API to enforce sandbox policies
• A kernel extension that provides support for regular expression pattern matching to enforce the defined policies

An application initializes the sandbox by calling sandbox_init. This function reads a human-friendly policy definition file and converts it into a binary format that is then passed to the kernel. Now the sandbox is initialized. Any function calls that are hooked by the TrustedBSD layer will be passed to the sandbox kernel extension for enforcement. Note that, unlike Janus, all enforcement takes place in the kernel. Enforcement means consulting a list of sandbox rules for the process that made the system call [the policy that was sent to the kernel by sandbox_init]. In some cases, the rules may involve regular expression pattern matching, such as those that define filename patterns].

The Apple sandbox helps avoid comprehension errors by providing predefined sandbox profiles [entitlements]. Certain resources are restricted by default and a sandboxed app must explicitly ask the user for permission. This includes accessing:

• the system hardware [camera, microphone, USB]
• network connections, data from other apps [calendar, contacts]
• location data, and user files [photos, movies, music, user-specified files]
• iCloud services

For mobile devices, there are also entitlements for push notifications and Apple Pay/Wallet access.

Once permission is granted, the sandbox policy can be modified for that application. Some basic categories of entitlements include:

• Restrict file system access: stay within an app container, a group container, any file in the system, or temporary/global places
• Deny file writing
• Deny networking
• Deny process execution

Browser-based Sandboxing: Chromium Native Client [NaCl]

Since the early days of the web, browsers have supported a plug-in architecture, where modules [containing native code] could be loaded into the browser to extend its capabilities. When a page specifies a specific plug-in via an or element, the requested content is downloaded and the plug-in that is associated with that object type is invoked on that content. Examples of common plug-ins include Adobe Flash, Adobe Reader [for rendering pdf files], and Java, but there are hundreds of others. The challenge with this framework is how to keep the software in a plug-in from doing bad things.

An example of sandboxing designed to address the problem of running code in a plugin is the Chromium Native Client,called NaCl. Chromium is the open source project behind the Google Chrome browser and Chrome OS. The NaCl Browser plug-in designed to allow safe execution of untrusted native code within a browser, unlike JavaScript, which is run through an interpreter. It is built with compute-intensive applications in mind or interactive applications that use the resources of a client, such as games.

NaCl is a user-level sandbox and works by restricting the type of code it can sandbox. It is designed for the safe execution of platform-independent, untrusted native code inside a browser. The motivation was that some browser-based applications will be so compute-intensive that writing them in JavaScript will not be sufficient. These native applications may be interactive and may use various client resources but will need to do so in a controlled and monitored manner.

NaCl supports two categories of code: trusted and untrusted. Trusted code can run without a sandbox. Untrusted code must run inside a sandbox. This code has to be compiled using the NaCl SDK or any compiler that adheres to NaCl’s data alignment rules and instruction restrictions [not all machine instructions can be used]. Since applications cannot access resources directly, the code is also linked with special NaCl libraries that provide access to system services, including the file system and network. NaCl includes a GNU-based toolchain that contains custom versions of gcc, binutils, gdb, and common libraries. This toolchain supports 32-bit ARM, 32-bit Intel x86 [IA–32], x86–64, and 32-bit MIPS architectures.

NaCl executes with two sandboxes in place:

1. The inner sandbox uses Intel’s IA–32 architecture’s segmentation capabilities to isolate memory regions among apps so that even if multiple apps run in the same process space, their memory is still isolated. Before executing an application, the NaCl loader applies static analysis on thecode to ensure that there is no attempt to use privileged instructions or create self-modifying code. It also attempts to detect security defects in the code.

2. The outer sandbox uses system call interposition to restrict the capabilities of apps at the system call level. Note that this is done completely at the user level via libraries rather than system call hooking.

Process virtual machine sandboxes: Java

A different type of sandbox is the Java Virtual Machine. The Java language was originally designed as a language for web applets, compiled Java programs that would get download and run dynamically upon fetching a web page. As such, confining how those applications run and what they can do was extremely important. Because the author of the application would not know what operating system or hardware architecture a client had, Java would compile to a hypothetical architecture called the Java Virtual Machine [JVM]. An interpreter on the client would simulate the JVM and process the instructions in the application. The Java sandbox has three parts to it:

The bytecode verifier verifies Java bytecodes before they are executed. It tries to ensure that the code looks like valid Java byte code with no attempts to circumvent access restrictions, convert data illegally, bypass array bounds, or forge pointers.

The class loader enforces restrictions on whether a program is allowed to load additional classes and that key parts of the runtime environment are not overwritten [e.g., the standard class libraries]. The class loader ensures that malicious code does not interfere with trusted code nad ensures that trusted class librares remain accessible and unmodified. It implements ASLR [Address Space Layout Randomization] by randomly laying out Runtime data areas [stacks, bytecodes, heap].

The security manager enforces the protection domain. It defines what actions are safe and which are not; it creates the boundaries of the sandbox and is consulted before any access to a resource is permitted. It is called at the time an application makes a call to specific methods so it can provide run-time verification of whether a program has been given rights to invoke the method, such as file I/O or network access. Any actions not allowed by the security policy result in a SecurityException being thrown. The security manager is the component that allows the user to restrict an application from accessing files or accessing the network, for example.A user can create a security policy file that enumerates what an application can or cannot do.

Java security is deceptively complex. After over twenty years of bugs one hopes that the truly dangerous ones have been fixed. Even though the Java language itself is pretty secure and provides dynamic memory management and array bounds checking, buffer overflows have been found in the underlying C support library, which has been buggy in general. Varying implementations of the JVM environment on different platforms make it unclear how secure any specific client will be. Moreover, Java supports the use of native methods, libraries that you can write in compiled languages such as C that interact with the operating system directly. These bypass the Java sandbox.

Malware

Malware is a term that refers to any malicious software that is unintentionally installed on a computer system. Malware can be distributed in various ways: viruses, worms, unintentional downloads, or trojan horses. It may spy on user actions and collect information on them [spyware], or present unwanted ads [adware]. It may disable components of the system or encrypt files, undoing its damage if the owner pays money [ransomware]. The software may sit dormant and wait for directives from some coordinator [a command and control server], who assembled an arsenal of hundreds of thousands of computers ready to do its bidding [for example, launch a distributed denial of service, DDoS, attack]. Some software might be legitimate but may contain backdoors – undocumented ways to allow an outsider to use that software to perform other operations on your system.

Malware Motivation

A saying often paraphrased from Sun Tzu’s The Art of War is “know your enemy.” In the case of malware, it helps to understand why someone would want to install malicious software on your computer. There are numerous reasons. Some are:

Steal account credentialsIf an attacker can obtain someone’s login and password credentials on one system, there is a good chance that those same credentials will work on other systems.EspionageAn attacker may have an interest in spying on the activities of a particular person. Traditionally, this would have been done through planting covert cameras and recording devices. Now it is often easier to accomplish the same results - and more - by installing software on the target’s computer. Such software is called spyware.Data theftAn attacker may target a person at a specific company [or a student taking a certain class] in an attempt to exfiltrate data of strategic interest. Alternatively, an attacker may target people anonymously, with the hope of obtaining information of value, such as credit card data or bank account information.SabotageThere’s vengeance or vandalism: the attacker may want to destroy a target’s content or devices.Host servicesAn attacker may need to harness computing, storage, or network resources. This can help hide the owner’s identity or amass a large collection of computers. An attacker can set up servers to host contraband data [e.g., stolen credit cards, login credentials, illegal material], send spam email on a large scale, mine cryptocurrency for free, or create a botnet for DDoS [distributed denial of service] attacks.Adware and ad clickingAn attacker may add software to a system or reconfigure a browser or hosts file to present unwanted advertising in the form of pop-up windows or banners. Additionally, the malware may redirect search requests or create click-throughs on ads that the user never wanted.RansomwareFinally, the attacker may want money directly. Ransomware installs software to encrypt files that will be [hopefully] decrypted if ransom is paid. The emergence of cryptocurrencies led to a huge increase in ransomware since they enabled anonymous payments.

Another saying paraphrased from The Art of War is “all warfare is based on deception.” This is also useful to consider with malware since it is most often installed willingly by the user of the system via some form of deception rather than through the exploitation of bugs in the system.

Malware Infiltration

There are several ways in which malware gets onto a system.

An attacker can exploit vulnerabilities in system services, particularly network services, to inject code that will download the malware. Zero-day vulnerabilities are particularly useful to attackers. These are bugs that have been discovered but not yet reported to the software vendor or the general public and hence are not fixed. They are typically known only to those who discovered and exploited the vulnerability … or sold the exploit.

As such, an attacker can be confident that the exploit will work on virtually all systems running the software and does not have to rely on targets who were not diligent enough to keep their systems patched. Ideally [for the attacker], the vulnerabilities will allow malware to run with elevated privileges so they can access all parts of a system or conceal itself more effectively.

Related to this are N-day vulnerabilities. N-day vulnerabilities are known vulnerabilities. Because they are known, developers have the opportunity to patch the code with a fix and IT administrators have the ability to apply a patch, shut off services, or put some detection mechanisms in place. However, all this does not happen instantly. Attackers have a period of time — N days — between the time that a vulnerability is disclosed and the time that most systems have been patched to avoid the vulnerability.

Malware might be installed unknowingly via infected removable media, most commonly USB flash drives [in earlier years, it would have been CDs or floppy disks].

By far the most common way that malware enters a system is via deception: the legitimate user of the system installed it unknowingly. This uses a social engineering attack to convince the user that it is in his or her interest to install the software. Social engineering is the art of manipulating, influencing, or deceiving a user into taking some action that is not in his/her or the organization’s best interest. Goal of social engineers is to obtain your trust and get you to divulge information or provide them with some form of access. In computers, social engineering refers to any techniques used by an adversary to trick you into disclosing information, opening a file, downloading an attachment, reading a message, or running a program.

Websites may offer downloads of “security” software, system “cleaner” software, or software “updates,” none of which will do their purported task. An attacker may convince a user to click on a URL in an email attachment or a web page. Software obtained from file sharing services are also excellent venues for distributing malware. A user may try to avoid spending $4,000 for an AutoCAD license or $240/year for an Adobe Illustrator license and turn to a file sharing site to download a patched copy or a crack for the software that bypasses license checks. Quite often, these downloads contain malware instead of the desired software [what do you expect - the user is trying to be a thief downloading software from thieves].

An attacker may search collections of stolen email addresses [or usernames] and passwords. Since people often use the same name and password on multiple systems, this can often give the attacker access to other websites on which the user has accounts. Accounts for banking sites are, of course, particularly valuable since they can be a direct conduit for transferring money.

Given login information about a user, an attacker can log onto the systems or services as the owner of the account and install malware, monitor the internal organization, and even send email [e.g., contact other employees or friends].

Any information the attacker can get about a user can help an attacker create a more convincing social attack. The term pretexting refers to using a concocted scenario to contact a user and get additional information [e.g., an attacker can pretend to be a caller from the IT department or a high-level manager from another location to try to extract information; with some rudimentary information, the attacker can mention some employee, department, or project names to sound like a true insider].

Types of Malware

Worms and viruses

A virus is software that attaches itself to another piece of software. It may also be content, such as scripts inside a Microsoft Word document or PDF file, that will be accessed and hence executed by some software. It may also be an email attachment that contains a document or software with the malware or a link to the malware.

It might even be a modification of the boot loader of a computer or the firmware on a flash drive. The key point is that it does not run as an independent process. A virus may spread automatically by trying to

A virus is executed because another program ran. Viruses are often spread by sharing files or software. On a computer, a virus may replicate itself onto other files or software to maximize its chance of spreading and reduce its chance of being removed.

A worm is conceptually similar in that it can do the same damage to the computer as a virus can. The distinction from a virus is that a worm runs as a standalone process while a virus requires a host program.

Worms and virus are both designed to propagate to other computers, although they may require human intervention to spread. In other cases, they can replicate themselves and spread to other systems automatically, exploiting weaknesses in software on those computers to allow themselves to infiltrate those machines. The popular use of both terms, worm and virus, has often blurred the distinctions between them. People often refer to any malware as a virus.

When using non-legitimate ways of getting into a system or elevating their privileges, attackers often try to find zero-day vulnerabilities. These are vulnerabilities [bugs or configuration errors] that have not been publicly reported, or are newly discovered, and hence are unpatched. They are referred to as “zero-day” because developers have zero days to fix the problem.

Virus and worm components

Viruses and worms contains three components:

Infection mechanismThe infection mechanism is the component of a worm or virus that enables it to spread. It can exploit software vulnerabilities to connect to other systems, patch certain files, or alter start-up scripts. PayloadThis is the malicious part of the virus and contains the code that does the actual harm to the system such as uploading personal information or deleting files. In some cases, the payload may be a generic service that contacts a command and control server from which it gets specific instructions on what to do [e.g., mine cryptocurrency, send spam, participate in a DDoS attack].TriggerThe trigger, also called a logic bomb, is code that is run whenever a file containing the virus is run. It makes the decision whether the payload should be executed. For example, some viruses may stay dormant until a particular date, number of runs, or upon getting directives from a command and control server.

Malware residence: where does it live?

File infector virus

A file infector virus is a virus that adds itself to an executable program. The virus patches the program so that, upon running, control will flow to the the virus code. Ideally, the code will install itself in some unused area of the file so that the file length will remain unchanged. A comparison of file sizes with the same programs on other systems will not reveal anything suspicious. When the virus runs, it will run the infector to decide whether to install itself on other files. The trigger will then decide whether the payload should be executed. If not, the program will appear to run normally.

Bootloader malware

Bootkits, also known as boot sector viruses, are malware that targets the booting process of a system. The malware has an infector that installs itself in the Master Boot Record [MBR] of a disk. In older BIOS-based PC systems, the first sector of the bootable storage device is read into memory and executed when the system boots, Normally, the code that is loaded is the boot loader that then loads the operating system. By infecting the master boot record, the virus can repeatedly re-infiltrate the operating system or files on the disk even if any malware on the system was previously detected and removed.

Boot sector viruses were common in the early days of PCs when users often booted off floppy disks and shared these disks. The virus would often use DOS commands to install itself onto other disks that it detects. Users on those systems had full administrative rights to modify any part of the system.

These viruses have diminished as attackers found more appealing targets. However, there is no reason that malware that attacks on the bootloader should not be considered to be a continued threat. 2011 saw the emergence of ransomware that modified the boot loader to prevent the operating system from booting unless a ransom was paid. In 2016, Petya Trojan ransomware was deployed, which also infects the MBR and encrypts disk contents.

Infected flash drives

In the early days of PCs, people would share content by passing around floppy disks. This became a means for viruses to spread, which could be planted in either the boot sector or in files. These days, people share USB flash drives the way they used to share floppies.

Autorun

In earlier Windows systems, Microsoft provided a feature called AutoRun. It was designed to make the CD [and, later, DVD and flash drive] experience better for users, particularly when using CDs for software installation. If the CD contained a file called autorun.inf, Windows would automatically execute a program identified within that file. While this made the experience of figuring out what to do after a CD is inserted easier for most users, it created a horrific security vulnerability: all that an adversary had to do was to get you to insert the media. Moreover, this functionality worked with any removable storage so that inserting a flash drive would automatically run a program defined within autorun.inf on the drive.

Microsoft eventually removed this capability from flash drives but some manufacturers created USB drives that emulated a CD drive to offer the “convenience” of AutoRun. Microsoft ultimately removed this functionality altogether in Windows 7. However, there are still old, unpatched versions of Windows out there that can be exploited with this vulnerability.

A similar problem occurs in the KDE framework. KDE is a desktop environment widely used on Linux systems. Malicious .desktop and .directory files can be created to run malicious code. Whenever the user uses the KDE file viewer to navigate to the directory where these files are stored, the code contained within these files will execute without any user interaction. This problem has not been fixed as of August 2019.

USB Firmware

The more insidious problem with USB flash drives now is unprotected firmware. A USB flash drive is a bunch of memory as well as firmware – embedded software on the chip. The firmware runs when you plug the drive into your computer. It identifies the drive as a USB storage device and manages the transferring of data. You don’t see this firmware and cannot tell if it has been changed. Because the firmware defines what the USB device is, modified firmware on the flash drive could present the drive as a keyboard and send a set of keyboard commands to the host system [for example, commands to open the terminal window and delete files].

A USB device can have multiple profiles associated with it and thus present itself as multiple devices, so the flash drive can tell the computer it is a keyboard but also a flash drive, so the user will still be able to use the device as a storage device. The firmware could also modify file contents as they pass between the USB storage device and host computer. The same attack can be user on other USB devices. For example, an ethernet adapter can redirect network messages to an attacker’s site.

Reprogramming the firmware has not been exploited by malware thus far, at least not in a widespread manner, but the vulnerability has been demonstrated and the source code to do this is freely and readily available.

Data leakage

The most common problem with flash drives is their portability and small size: they are easy to lose and easy to borrow. This makes them vulnerable to data leakage, which is just a fancy term that means some adversary may access your data simply by borrowing your flash drive.

In 2016, researchers at the University of Illinois ran an experiment where they scattered nearly 300 USB drives in public areas through the campus. Each of those drives was loaded with files that, when opened on a network-connected computer, would contact a server to tell it that the drive has been picked up and the file was opened. The results of the study showed that 98% of the drives were picked up and someone opened up at least one file on 45% of them[1]. Because of the risk of malicious firmware, even formatting a drive does not make it safe to use.

Inadvertent program execution

The portability of flash drives makes them a distribution mechanism. Experiments of scattering a number of them in parking lots revealed that many people are all too willing to plug a random drive into their system.

Even without automatic execution capabilities enabled, attackers can use flash drives as a distribution mechanism for malware. The Stuxnet attack exploited a windows bug in rendering shortcut icons where just viewing them in Windows Explorer enabled the execution of arbitrary code. Others have exploited a bug in video playback that allowed code execution. Even something as simple as an HTML file on a drive may direct the target to a website that can launch an attack.

There are many other USB device-based attacks. Take a look here if you’re curious.

Macro viruses

Some applications have support for macros, which allow the user to define a set of commands to avoid repetitive tasks and improve productivity. They are particularly common in text editors but are present in other applications as well, such as Photoshop and Microsoft Word and Excel. In some cases, as with Microsoft Office applications, macros are embedded in the document, which means they can be passed on to other users who access that document. Some macro capabilities are far more powerful than simply defining repetitive commands. Microsoft Office products, for example, provide Visual Basic scripting, which effectively allows users to embed complete programs into their documents. VBScript is based on Visual Basic and provides features that make it easy to access network printers, network files, special folders, user information, and even execute scripts on remote systems.

Scripts in Office documents can spread not only by having the user pass the original infected document around but by modifying the default template file, normal.dot. This will affect every other document on the system. With operating systems providing better access controls and users not running with administrative privileges, embedded scripts are a ripe area for attack. If you can convince somebody to open a document, they will run your program on their machine.

The challenge, of course, is to get a file with a malicious macro to target users and get them to open it. One of the most common techniques is to send it as an email attachment with some inducement to get the user to click on the document. This is an example of social engineering.

One hugely-successful virus that did this was the ILOVEYOU virus from 2000. The subject of the message stated that it is a letter from a secret admirer. The attachment wasn’t even a document; it was a visual basic script. To provide a better user experience, Microsoft would hide file extensions by default [macOS does this too]. The file was named LOVE-LETTER-FOR-YOU.TXT.vbs but the .vbs suffix, which indicated that the file was a visual basic script, was hidden from users, so they only saw LOVE-LETTER-FOR-YOU.TXT. Not being aware of when extensions are hidden and when they are not, millions of users assumed they received an innocuous text file and clicked on it. Upon execution, the script would copy itself into various folders, modify and add new entries to the system registry, replace various types of files with copies of itself [targeting music and video files], and try to propagate itself through Internet relay Chat clients as well as email. If that wasn’t enough, it would download a file called WIN-BUGFIX.EXE and execute it. This was not a bug fixing program but rather a program that extracted user passwords and mailed them to the hacker.

The ILOVEYOU virus transmitted itself largely through email to contacts in infected computers, so your “secret admirer” message came from someone you knew and hence you were more likely to click on it. An earlier highly successful virus, Melissa, spread by offering a list of passwords for X-rated web sites. Email-based virus transmission is still a dominant mechanism. Sender headers and links are often disguised to make it look like the content is from a legitimate party.

JavaScript and PDF files

JavaScript, like Visual Basic, has evolved into a full programming language. Most browsers have security holes that involve Javascript. JavaScript can not only modify the content and structure of a web page but can connect to other sites. This allows any malicious site to leverage your machine. For example, systems can perform port scans on a range of IP addresses and report any detected unsecured services.

PDF [Portable Document Format] files, would seem to be innocent printable documents, incapable of harboring executable code. However, PDF is a complex format that can contain a mix of static and dynamic elements. Dynamic elements may contain Javascript, dynamic action triggers [e.g., “on open”], and the ability to retrieve “live” data via embedded URLs. As with Visual Basic scripts, PDF readers warn users of dynamic content but, depending on the social engineering around the file, the user may choose to trust the file … or not even pay attention to the warning in yet-another-dialog-box.

Trojans

A Trojan horse is a program with two purposes: an overt purpose and a covert one. The overt purpose is what compels the user to get and run the program in the first place. The covert purpose is unknown to the user and is the malicious part of the program.

For example, a script with the name of a common Linux command might be added to a target user’s search path. When the user runs the command, the script is run. That script may, in turn, execute the proper command, leading the user to believe that all is well. As a side effect, the script may create a setuid shell to allow the attacker to impersonate that user or mail copy over some critical data. Users install Trojans because they believe they are installing useful software, such as an anti-virus tool [BTW, a lot of downloadable hacker tools contain Trojans: hackers hacking wannabe hackers]. The side-effect of this software can activate cameras, enable key loggers, or deploy bots for anonymization servers, DDoS attacks, or spam attacks.

Trojans may include programs [games, utilities, anti-malware programs], downloading services, rootkits [see next] and backdoors [see next]. They appear to perform a useful task that does not raise suspicion on the part of the victim.

Backdoors

A backdoor is software that is designed with some undocumented mechanism to allow someone who knows about it to be able to access the system or specific functions in a way that bypasses proper authentication mechanisms. In many cases, they are not designed for malicious use: they may allow a manufacturer to troubleshoot a device or a software author to push an update. However, if adversarial parties discover the presence of a backdoor, they can use it for malicious purposes.

An old, but famous, example of a backdoor is the sendmail mail delivery server. The author of sendmail wanted to have development access on a production system that had the program installed so that he can continue to improve it. The system administrator refused such access. His next release of sendmail contained a password-protected backdoor that gave him access to the system via the sendmail server. The password was hard-coded in the program and soon became well-known. Robert Morris used the knowledge of this backdoor as one of the mechanisms for his worm to propagate to other systems. More recently, in 2014, some Samsung Galaxy phones were delivered with backdoors that provide remote access to the data on the phone.

Rootkits

A rootkit is software that is designed to allow an attacker to access a computer and hide the existence of the software … and sometimes hide the presence of the user on the system.

Historically, a basic rootkit would replace common administration commands [such as ps, ls, find, top, netstat, etc.] with commands that mimic their operation but hide the presence of intruding users, intruding processes, and intruding files. The idea is that a system administrator should be able to examine the system and believe that all is fine and the system is free of malware [or of unknown user accounts].

User mode rootkits

The rootkit just described is a user mode rootkit and involves replacing commands, intercepting messages, and patching commonly-used APIs that may divulge the presence of the malware. A skilled administrator may find unmodified commands or import software to detect the intruding software.

Kernel mode rootkits

A kernel mode rootkit is installed as a kernel module. Being in the kernel gives the rootkit unrestricted access to all system resources and the ability to patch kernel structures and system calls. For example, directory listings from the getdents64 system call may not report any names that match the malware. Commands and libraries can be replaced and not give any indication that malicious software is resident in the system.

Hypervisor rootkits

The most insidious rootkits are hypervisor rootkits. A hypervisor sits below the operating system and is responsible for translating between virtual device operations from operating systems and the underlying hardware. All I/O flows through the hypervisor. Most computer systems do not run virtual machines and hence have no hypervisor. These systems are prime targets for a hypervisor-based rootkit. Now you can have an environment where the entire operating system can run unmodified - or even be reinstalled - and be unaware that its operations are being intercepted at a lower level. The hypervisor does not have to virtualize all hardware interactions: just the ones it cares about. For example, it might want to grab keyboard events to record passwords and messages.

Hypervisor attacks have not been deployed but have been demonstrated as a proof of concept. The challenge in detecting their presence is that operating systems are unaware if they are running under a hypervisor, so if a malicious hypervisor is installed, the operating system needs to detect that it is running under a hypervisor rather than directly on the computer system. Detection is difficult and often relies on measuring completion times of certain system calls. If they go through a hypervisor, they will take a longer time and the on-chip Time Stamp Counter [TSC], which counts CPU cycles, will show a longer value with a hypervisor in place. An alternative, and far more obscure, method of detection, is the use of an instruction that stores the interrupt descriptor table register [IDTR] into a memory location [the SIDT instruction]. The hypervisor changes the register’s value and the instruction can detect that. However, this does not have to take place on a system with only one operating system, so measuring timing differences may still be the more foolproof approach.

Ransomware

If we think back to the goals of malware, one common goal was to extract money: even hackers need to monetize their efforts. An indirect way of accomplishing this was by collecting information to gain access to bank account data, PayPal data, or modifying accounts that may take money, such as eBay accounts. A more direct way of getting money is to demand it from the victim. Ransomware is a relatively new form of malware that locks a computer, keeps it from booting, or encrypts all files on the system. It then asks the suer to pay a ransom [usually via bitcoin] to get a decryption program.

Gathering information

Malware has varying goals. These goals may include spying on user activity, destroying content, assembling a collection of servers, or extracting money from a victim. One common goal is to gather information … or get the user to provide information. Your computer might not have anything of direct value to an adversary, but your PayPal, bank, Amazon, or eBay credentials might be useful.

Phishing

Phishing is a social engineering attack whose most common purpose is to get personal information from someone, usually login credentials to some service. These are often carried out vie email with similar techniques that are used to spread infected files. A message announcing that your PayPal account is being canceled, that your bank detected a fraudulent transaction, or that FedEx could not deliver a package may prompt the receiver to panic and immediately click on a link in the message, which may result in the browser displaying a site crafted to look like PayPal, the bank, or FedEx and prompt the user for login and password information.

Phishing attacks are surprisingly effective. A 2018 study by Proofpoint found that 52% of all successful phishing emails are clicked on within one hour of being sent.

Spear phishing is a targeted form of phishing. A phishing attack sends the same message to a large set of users, hoping that some percentage of them will be fooled. A spear phishing attack sends a customized message that demonstrates some knowledge of the target, which will usually lead the target to think that the message is legitimate. For example, the 2016 Democratic National Committee [DNC] was facilitated by spear phishing. Targets were sent a message containing bit.ly links, which is a common URL shortening service that hid the actual underlying URLs. Once clicked, the web site would display what looked like a legitimate Google accounts login page, already pre-populated with the victim’s GMail address.

More recent GMail spear phishing attacks send email to contacts of compromised accounts. The email contains an innocent-looking attachment: a thumbnail image of a document. When the victim clicks on the attachment, a web page that looks like a legitimate Google sign-in page is presented. As soon as the victim enters a name and password, the attackers get the credentials, log into the account, and target people in the victim’s contact list. They use an image of an actual attachment in the victim’s email and an actual subject line to make the email look more legitimate.

A 2017 report by Webroot found that 1.385 million new and unique phishing sites are created each month. Some warning signs that a mail message may be a phishing attack are:

1. From header: is it from an unknown or suspicious address?

2. To header: if the message is sent to multiple people, do you recognize any other names on the header?

3. Date header: if the message purports to be a personal message, was it sent during normal business hours?

4. Subject header: is the suspicious and is it relevant to your activities?

5. Message content: is the message a request to click on a link in order to avoid a negative consequence?

6. Embedded links: are there any links that you are asked to click? If you look at the target of those links, are they misspelled, suspicious, or for a site different from that of the sender?

7. Attachments: is there an unexpected attachment that you are expected to open, such as a Microsoft Word document or PDF file?

Deceptive web sites

Quite often, malicious links in phishing attacks direct the user to a web site in order to obtain their login credentials. These sites masquerade as legitimate sites. The Proofpoint study mentioned earlier found that for every legitimate website, there are 20 malicious sites that mimic it. This is known as typosquatting. Such sites can be masqueraded banking sites, Google/Microsoft/Apple authentication pages, videoconferencing plugin-software downloads, etc.

File serving sites, including those that host software or those that provide services such as PDF or mp3 conversion are often ad-sponsored. Some of the ads on these sites, however, often look like download links and can trick a user into clicking on the ad instead of the link for the actual content. The

Keyloggers

Another way of obtaining information is to snoop on a user’s actions. Keyloggers record everything a victim types and allow a user to extract login names, passwords, and entire messages.

Keyloggers can be implemented in several ways:

Malicious hypervisorSince a hypervisor provides virtual interfaces for all the resources of a computer, it can capture all keyboard, mouse, and even video data. These attacks are difficult since they rely on the ability to install a hypervisor.Kernel-based rootkit All input/output operations go through the operating system kernel. Modifying the kernel allows malicious software to log and upload keystroke data.System call hookingSome operating systems provide a system call hooking mechanism that allows data to and from system calls to be intercepted. We saw how this was used to implement sandboxing. Windows enables this without having to install any kernel-level drivers. The SetWindowsHookEx system call can be used to report WH_KEYBOARD and WH_MOUSE events, capturing keyboard and mouse activity.Browser-based loggingJavaScript can be used to capture onKeyUp[] events. These events will be captured for one page but other hacks can be used to create a broader context with embedded pages. Form submission can also be intercepted to get populated form data without having to reassemble key presses into coherent account credentials.Hardware loggersAlthough visible to the user, hardware key loggers can be used for USB-connected keyboards. Some of these have embedded Wi-Fi transceivers that enable an attacker to collect the data from a distance.

Defenses

Malware was particularly easy to spread on older Windows systems since user accounts, and hence processes, ran with full administrative rights, which made it easy to modify any files on the system and even install kernel drivers. Adding file protection mechanisms, such as a distinction between user and administrator accounts added a significant layer of protection. However, malware installed by the user would run with that user’s privileges and would have full access to all of a user’s files. If any files are read or write protected, the malware can change DAC permissions.

Systems took the approach of warning users if software wanted to install software or asked for elevated privileges. Social engineering hopes to convince users that they actually want to install the software [or view the document]. They will happily grant permissions and install the malware. MAC permissions can stop some viruses as they will not be able, for instance, to override write permissions on executable files but macro viruses and the user files are still a problem.

In general, however, studies have shown that by simply taking away admin rights [avoiding privilege escalation] from users, 94% of the 530 Microsoft vulnerabilities that were reported in 2016 could be mitigated and 100% of vulnerabilities in Office 2016 could be mitigated.

Anti-virus [anti-malware] software

There is no way to recognize all possible viruses. Anti-virus software uses two strategies: signature-based and behavior-based approaches.

With signature-based systems, anti-virus programs look for byte sequences that match those in known malware. Each bit pattern is an excerpt of code from a known virus and is called a signature [not to be confused with digital signatures, discussed later in the course]. A virus signature is simply a set of bytes that make up a portion of the virus and allow scanning software to see whether that virus is embedded in a file. The hope is that the signature is long enough and unique enough that the byte pattern will not occur in legitimate programs. This scanning process is called signature scanning. Lists of signatures [“virus definitions”] have to be updated by the anti-virus software vendor as new viruses are discovered. Signature-based detection is used by most anti-virus products.

A behavior-based system monitors the activities of a process [typically the system calls or standard library calls that it makes]. Ideally, sandboxing is employed, to ensure that the suspected code is run within a sandbox or even in an interpreted environment within a sandbox to ensure that it cannot cause real damage. Behavior-based systems try to perform anomaly detection. If the observed activity is deemed suspicious, the process is terminated and the user alerted. Sandboxed, behavior-based analysis is often run by anti-malware companies to examine what a piece of suspected malware is actually doing and whether it should be considered to be a virus. A behavior-based can identify previously-unseen malware but these systems tend to have higher false positive rates of detection: it is difficult to characterize exactly what set of operations constitute suspicious behavior.

Windows Defender, as an example, makes use of both signature-based scanning as well as behavior-based process monitoring. It uses signature-based scanning on files and behavior-based analysis for running processes. Behavior monitoring includes scanning for suspicious file and registry changes [e.g., ransomware may try to encrypt all the files on a system and a lot of malware may try to modify the registry so that the software runs when the system is rebooted.

Countermeasures

Some viruses will take measures to try to defend themselves from anti-virus software.

Signature scanning countermeasures

A common thing to do in malware is to use a packer on the code, unpacking it prior to execution. Packing can be one of several operations:

• Simply obscure the malware payload by exclusive-oring [xor] with a repeating byte pattern [exclusive-oring the data with the same byte pattern reconstructs it.
• Compress the code and then uncompress it upon loading it prior to execution.
• Encrypt the code and decrypt it prior to execution.

All of these techniques will change the signature of a virus. One can scan for a signature of a compressed version of the virus but there are dozens of compression algorithms around, so the scanning process gets more complicated.

With encryption [xor is a simple form of encryption], only the non-encrypted part of the virus contains the unpacking software [decryption software and the key]. A virus scanner will need to match the code for the unpacker component since the key and the encrypted components can change each time the virus propagates itself.

Polymorphic viruses mutate their code each time they run while keeping the algorithm the same. This involves replacing sequences of instructions with functionally-identical ones. For example, one can change additions to subtractions of negative numbers, invert conditional tests and branches, and insert or remove no-op instructions. This thwarts signature scanning software because the the byte pattern of the virus is different each time.

Access control countermeasures

Access controls help but do not stop the problem of malware. Containment mechanisms such as containers work well for server software but are usually impractical for user software [e.g., you want Microsoft Word to be able to read documents anywhere in a user’s directories]. Application sandboxing is generally far more effective and is a dominant technique used in mobile software.

Trojans, deceptive downloads, and phishing attacks are insidiously difficult to defend against since we are dealing with human nature: users want to install the software or provide the data. They are conditioned to accepting pop-up messages and entering a password. Better detection in browsers & mail clients against suspicious content or URLs helps. However, malware distributors have been known to simply ask a user to rename a file to turn it into one that is recognized by the operating system as an executable file [or a disk image, PDF, or whatever format the malware come in and may otherwise be filtered by the mail server or web browser.

Sandboxing countermeasures

Virusus are unlikely to get through a sandbox [unless there are vulnerabilities or an improper configuration]. However, there are areas where malware can address sandboxing:

1. Vendor examination
   Anti-virus vendors often test software within a tightly configured sandboxed environment so they can detect whether the software is doing anything malicious [e.g., accessing files, devices, or the network in ways it is not supposed to]. If they detect that they do have malware, they will dig in further and extract a signature so they can update and distribute their list of virus definitions. Viruses can try to get through this examination phase by setting a trigger to keep the virus from immediately performing malicious actions or to stay dormant for the first several invocations. The hope is that the anti-virus vendors will not see anything suspicious and the virus will never be flagged as such by their software.

2. User configuration [entitlements]
   Virtually all mobile applications, and increasingly more desktop/laptop applications, are run with application sandboxes in place. These may disallow malware from accessing files, devices, or the network. However, it never hurts to ask. The software can simply ask the user to modify the sandbox settings. If social engineering is successful, the user may not even be suspicious and not wonder why a game wants access to contacts or location information.

Cryptography

Cryptography deals with encrypting plaintext using a cipher, also known as an encryption algorithm, to create ciphertext, which is unintelligible to anyone unless they can decrypt the ciphertext. It is a tool that helps build protocols that address:

AuthenticationShowing that the user really is that user.Integrity: Validating that the message has not been modified.Nonrepudiation:Binding the origin of a message to a user so that she cannot deny creating it.Confidentiality:Hiding the contents of a message.

A restricted cipher is one where the workings of the cipher must be kept secret. There is no reliance on any key and the secrecy of the cipher is crucial to the value of the algorithm. This has obvious flaws: people in the know leaking the secret, designers coming up with a poor algorithm, and reverse engineering.

For any serious use of encryption, we use well-tested, non-secret algorithms that rely on secret keys. A key is a parameter to a cipher that alters the resulting ciphertext. Knowledge of the key is needed to decrypt the ciphertext. Kerckhoffs’s Principle states that a cryptosystem should be secure even if everything about the system, except the key, is public knowledge. We expect algorithms to be publicly known and all security to rest entirely on the secrecy of the key.

A symmetric encryption algorithm uses the same secret key for encryption and decryption.

An alternative to symmetric ciphers are asymmetric ciphers. An asymmetric, or public key cipher uses two related keys. Data encrypted with one key can only be decrypted with the other key.

Properties of good ciphers

For a cipher to be considered good, ciphertext should be indistinguishable from random values. Given ciphertext, there should be no way to extract the original plaintext or the key that was used to create it except by of enumerating over all possible keys. This is called a brute-force attack. The keys used for encryption should be large enough that a brute force attack is not feasible. Each additional bit in a key doubles the number of possible keys and hence doubles the search time.

Classic cryptography

Monoalphabetic substitution ciphers

The earliest form of cryptography was the monoalphabetic substitution cipher. In this cipher, each character of plaintext is substituted with a character of ciphertext based on a substitution alphabet [a lookup table]. The simplest of these is the Caesar cipher, known as a shift cipher, in which a plaintext character is replaced with a character that is n positions away in the alphabet. The key is the simply the the shift value: the number n. Substitution ciphers are vulnerable to frequency analysis attacks, in which an analyst analyzes letter frequencies in ciphertext and substitutes characters with those that occur with the same frequency in natural language text [e.g., if “x” occurs 12% of the time, it’s likely to really be an “e” since “e” occurs in English text approximately 12% of the time while “x” occurs only 0.1% of the time].

Polyalphabetic substitution ciphers

Polyalphabetic substitution ciphers were designed to increase resiliency against frequency analysis attacks. Instead of using a single plaintext to ciphertext mapping for the entire message, the substitution alphabet may change periodically. Leon Battista Alberti is credited with creating the first polyalphabetic substitution cipher. In the Alberti cipher [essentially a secret decoder ring], the substitution alphabet changes every n characters as the ring is rotated one position every n characters.

The Vigenère cipher is a grid of Caesar ciphers that uses a repeating key. A repeating key is a key that repeats itself for as long as the message. Each character of the key determines which Caesar cipher [which row of the grid] will be used for the next character of plaintext. The position of the plaintext character identifies the column of the grid. These algorithms are still vulnerable to frequency analysis attacks but require substantially more plaintext since one needs to deduce the key length [or the frequency at which the substitution alphabet changes] and then effectively decode multiple monoalphabetic substitution ciphers.

One-time Pads

The one-time pad is the only provably secure cipher. It uses a random key that is as long as the plaintext. Each character of plaintext is permuted by a character of ciphertext [e.g., add the characters modulo the size of the alphabet or, in the case of binary data, exclusive-or the next byte of the text with the next byte of the key]. The reason this cryptosystem is not particularly useful is because the key has to be as long as the message, so transporting the key securely becomes a problem. The challenge of sending a message securely is now replaced with the challenge of sending the key securely. The position in the key [pad] must by synchronized at all times. Error recovery from unsynchronized keys is not possible. Finally, for the cipher to be secure, a key must be composed of truly random characters, not ones derived by an algorithmic pseudorandom number generator. The key can never be reused.

The one-time pad provides perfect secrecy [not to be confused with forward secrecy, also called perfect forward secrecy, which will be discussed later], which means that the ciphertext conveys no information about the content of the plaintext. It has been proved that perfect secrecy can be achieved only if there are as many possible keys as the plaintext, meaning the key has to be as long as the message. Watch this video for an explanation of perfect secrecy.

Stream ciphers

A stream cipher simulates a one-time pad by using a keystream generator to create a set of key bytes that is as long as the message. A keystream generator is a pseudorandom number generator that is seeded, or initialized, with a key that drives the output of all the bytes that the generator spits out. The keystream generator is fully deterministic: the same key will produce the same stream of output bytes each time. Because of this, receivers only need to have the key to be able to decipher a message. However, because the keystream generator does not generate true random numbers, the stream cipher is not a true substitute for a one-time pad. Its strength rests on the strength of the key. A keystream generator will, at some point, will reach an internal state that is identical to some previous internal state and produce output that is a repetition of previous output. This also limits the security of a stream cipher but the repetition may not occur for a long time, so stream ciphers can still be useful for many purposes.

Rotor machines

A rotor machine is an electromechanical device that implements a polyalphabetic substitution cipher. It uses a set of disks [rotors], each of which implements a substitution cipher. The rotors rotate with each character in the style of an odometer: after a complete rotation of one rotor, the next rotor advances one position. Each successive character gets a new substitution alphabet applied to it. The multi-rotor mechanism allows for a huge number of substitution alphabets to be employed before they start repeating when the rotors all reach their starting position. The number of alphabets is cr, where c is the number of characters in the alphabet and r is the number of rotors.

Transposition ciphers

Instead of substituting one character of plaintext for a character of ciphertext, a transposition cipher scrambles the position of the plaintext characters. Decryption is the knowledge of how to unscramble them.

A scytale, also known as a staff cipher, is an ancient implementation of a transposition cipher where text written along a strip of paper is wrapped around a rod and the resulting sequences of text are read horizontally. This is equivalent to entering characters in a two-dimensional matrix horizontally and reading them vertically. Because the number of characters might not be a multiple of the width of the matrix, extra characters might need to be added at the end. This is called padding and is essential for block ciphers, which encrypt chunks of data at a time.

Block ciphers

Most modern ciphers are block ciphers, meaning that they encrypt a chunk of bits, or block, of plaintext at a time. The same key is used to encrypt each successive block of plaintext.

AES and DES are two popular symmetric block ciphers. Symmetric block ciphers are usually implemented as iterative ciphers. The encryption of each block of plaintext iterates over several rounds. Each round uses a subkey, which is a key generated from the main key via a specific set of bit replications, inversions, and transpositions. The subkey is also known as a round key since it is applied to only one round, or iteration. This subkey determines what happens to the block of plaintext as it goes through a substitution-permutation [SP] network. The SP network, guided by the subkey, flips some bits by doing a substitution, which is a table lookup of an input bit pattern to get an output bit pattern and a permutation, which is a scrambling of bits in a specific order. The output bytes are fed into the next round, which applies a substitution-permutation step onto a different subkey. The process continues for several rounds [16 rounds for DES, 10–14 rounds for AES]. and the resulting bytes are the ciphertext for the input block.

The iteration through multiple SP steps creates confusion and diffusion. Confusion means that it is extremely difficult to find any correlation between a bit of the ciphertext with any part of the key or the plaintext. A core component of block ciphers is the s-box, which converts n input bits to m output bits, usually via a table lookup. The purpose of the s-box is to add confusion by altering the relationship between the input and output bits.

Diffusion means that any changes to the plaintext are distributed [diffused] throughout the ciphertext so that, on average, half of the bits of the ciphertext would change if even one bit of plaintext is changed.

Feistel ciphers

A Feistel cipher is a form of block cipher that uses a variation of the SP network where a block plaintext is split into two parts. The substitution-permutation round is applied to only one part. That output is then XORed with the other part and the two halves are swapped. At each round, half of the input block remains unchanged. DES, the Data Encryption Standard, is an example of a Feistel cipher. AES, the Advanced Encryption Standard, is not.

DES

Two popular symmetric block ciphers are DES, the Data Encryption Standard, and AES, the Advanced Encryption Standard. DES was adopted as a federal standard in 1976 and is a block cipher based on the Feistel cipher that encrypts 64-bit blocks using a 56-bit key.

DES has been shown to have some minor weaknesses against cryptanalysis. Key can be recovered using 247 chosen plaintexts or 243 known plaintexts. Note that this is not a practical amount of data to get for a real attack. The real weakness of DES is not the algorithm but but its 56-bit key. An exhaustive search requires 255 iterations on average [we assume that, on average, the plaintext is recovered halfway through the search]. This was a lot for computers in the 1970s but is not much for today’s dedicated hardware or distributed efforts.

Triple-DES

Triple-DES [3DES] solves the key size problem of DES and allows DES to use keys up to 168 bits. It does this by applying three layers of encryption:

1. C’ = Encrypt M with key K1
2. C’’ = Decrypt C’ with key K2
3. C = Encrypt C’’ with key K3

If K1, K2, and K3 are identical, we have the original DES algorithm since the decryption in the second step cancels out the encryption in the first step. If K1 and K3 are the same, we effectively have a 112-bit key and if all three keys are different, we have a 168-bit key.

Cryptanalysis is not effective with 3DES: the three layers of encryption use 48 rounds instead of 16 making it infeasible to reconstruct the substitutions and permutations that take place. DES is relatively slow compared with other symmetric ciphers, such as AES. It was designed with hardware encryption in mind 3DES is, of course, three times slower than DES.

AES

AES, the Advanced Encryption Standard, was designed as a successor to DES and became a federal government standard in 2002. It uses a larger block size than DES: 128 bits versus DES’s 64 bits and supports larger key sizes: 128, 192, and 256 bits. Even 128 bits is complex enough to prevent brute-force searches.

No significant academic attacks have been found thus far beyond brute force search. AES is also typically 5–10 times faster in software than 3DES.

Block cipher modes

Electronic Code Book [ECB]

When data is encrypted with a block cipher, it is broken into blocks and each block is encrypted separately. This leads to two problems.

1. If different encrypted messages contain the same substrings and use the same key, an intruder can deduce that it is the same data.

2. Secondly, a malicious party can delete, add, or replace blocks [perhaps with blocks that were captured from previous messages].

This basic form of a block cipher is called an electronic code book [ECB]. Think of the code book as a database of encrypted content. You can look up a block of plaintext and find the corresponding ciphertext. This is not feasible to implement for arbitrary messages but refers to the historic use of codebooks to convert plaintext messages to ciphertext.

Cipher Block Chaining [CBC]

Cipher block chaining [CBC] addresses these problems. Every block of data is still encrypted with the same key. However, prior to being encrypted, the data block is exclusive-ORed with the previous block of ciphertext. The receiver does the process in reverse: a block of received data is decrypted and then exclusive-ored with the previously-received block of ciphertext to obtain the original data. The very first block is exclusive-ored with a random initialization vector, which must be transmitted to the remote side.

Note that CBC does not make the encryption more secure; it simply makes the result of each block of data dependent on all previous previous blocks so that data cannot be meaningfully inserted or deleted in the message stream.

Counter mode [CTR]

Counter mode [CTR] also addresses these problems but in a different way. The ciphertext of each block is a function of its position in the message. Encryption starts with a message counter. The counter is incremented for each block of input. Only the counter is encrypted. The resulting ciphertext is then exclusive-ORed with the corresponding block of plaintext, producing a block of message ciphertext. To decrypt, the receiver does the same thing and needs to know the starting value of the counter as well as the key.

An advantage of CTR mode is that each block has no dependance on other blocks and encryption on multiple blocks can be done in parallel.

Cryptanalysis

The goal of cryptanalysis is break codes. Most often, it is to identify some non-random behavior of an algorithm that will give the analyst an advantage over an exhaustive search of the key space.

Differential cryptanalysis seeks to identify non-random behavior by examining how changes in plaintext input affect changes in the output ciphertext. It tries to find whether certain bit patterns are unlikely for certain keys or whether the change in plaintext results in likely changes in the output.

Linear cryptanalysis tries to create equations that attempt to predict the relationships between ciphertext, plaintext, and the key. An equation will never be equivalent to a cipher but any correlation of bit patterns give the analyst an advantage.

Neither of these methods will break a code directly but may help find keys or data that are more likely are that are unlikely. It reduces the keys that need to be searched.

Public key cryptography

Public key algorithm, also known as asymmetric ciphers, use one key for encryption and another key for decryption. One of these keys is kept private [known only to the creator] and is known as the private key. The corresponding key is generally made visible to others and is known as the public key.

Anything encrypted with the private key can only be decrypted with the public key. This is the basis for digital signatures. Anything that is encrypted with a public key can be encrypted only with the corresponding private key. This is the basis for authentication and covert communication.

Public and private keys are related but, given one of the keys, there is no feasible way of computing the other. They are based on trapdoor functions, which are one-way functions: there is no known way to compute the inverse unless you have extra data: the other key.

RSA public key cryptography

The RSA algorithm is the most popular algorithm for asymmetric cryptography. Its security is based on the difficulty of finding the factors of the product of two large prime numbers. Unlike symmetric ciphers, RSA encryption is a matter of performing arithmetic on large numbers. It is also a block cipher and plaintext is converted to ciphertext by the formula:

c = me mod n

Where m is a block of plaintext, e is the encryption key, and n is an agreed-upon modulus that is the product of two primes. To decrypt the ciphertext, you need the decryption key, d:

m = cd mod n

Given the ciphertext c, e, and n, there is no efficient way to compute the inverse to obtain m. Should an attacker find a way to factor n into its two prime factors, however, the attacker would be able to reconstruct the encryption and decryption keys, e and d.

Elliptic curve cryptography [ECC]

Elliptic curve cryptography [ECC] is a more recent public key algorithm that is an alternative to RSA. It is based on finding points along a prescribed elliptic curve, which is an equation of the form:

y2 = x3 + ax + b

Contrary to its name, elliptic curves have nothing to do with ellipses or conic sections and look like bumpy lines. With elliptic curves, multiplying a point on a given elliptic curve by a number will produce another point on the curve. However, given that result, it is difficult to find what number was used. The security in ECC rests not our inability to factor numbers but our inability to perform discrete logarithms in a finite field.

The RSA algorithm is still the most widely used public key algorithm, but ECC has some advantages:

• ECC can use far shorter keys for the same degree of security. Security comparable to 256 bit AES encryption requires a 512-bit ECC key but a 15,360-bit RSA key

• ECC requires less CPU consumption and uses less memory than RSA. It is faster for encryption [including signature generation] than RSA but slower for decryption.

• Generating ECC keys is faster than RSA [but much slower than AES, where a key is just a random number].

On the downside, ECC is more complex to implement and decryption is slower than with RSA. As a standard, ECC was also tainted because the NSA inserted weaknesses into the ECC random number generator that effectively created a backdoor for decrypting content. This has been remedied and ECC is generally considered the preferred choice over RSA for most applications.

If you are interested, see here for a somewhat easy-to-understand tutorial on ECC.

Quantum computing

Quantum computers are a markedly different form computer. Conventional computers store and process information that is represented in bits, with each bit having a distinct value of 0 or 1. Quantum computers use the principles of quantum mechanics, which include superposition and entanglement. Instead of working with bits, quantum computers operate on qubits, which can hold values of “0” and “1” simultaneously via superposiion. The superpositions of qubits can be entangled with other objects so that their final outcomes will be mathematically related. A single operation can be carried out on 2^n values simultaneously, where n is the number of qubits in the computer.

be solved exponentially faster than with conventional comptuers. Shor’s algorithm, for instance, will be able to find the prime factors of large integers and compute discrete logarithms far more efficiently than is currently possible.

So far, quantum computers are very much in their infancy and it is not clear when – or if – large-scale quantum computers that are capable of solving useful problems will be built. IBM and Google are two companies that are racing to build one. It is unlikely that they will be built in the next several years but we expect that they will be built eventually. Shor’s algorithm will be able to crack public-key based systems such as RSA, Elliptic Curve Cryptography, and Diffie-Hellman key exchange. In 2016, the NSA called for a migration to “post-quantum cryptographic algorithms” and has currently narrowed down the submissions to 26 candidates. The goal is to find useful trapdoor functions that do not rely on multiplying large primes, computing exponents, any other mechanisms that can be attacked by quantum computation. If you are interested in these, you can read the NSA’s report.

Symmetric cryptosystems, such as AES, are not particularly vulnerable to quantum computing since they rely on moving and flipping bits rather than applying mathematical functions on the data. The best potential attacks come via Grover’s algorithm, which yields only a quadratic rather than an exponential speedup in key searches. This will reduce the effective strength of a key by a factor of two. For instance, a 128-bit key will have the strength of a 64-bit key on a conventional computer. It is easy enough to use a sufficiently long key [256-bit AES keys are currently recommended] so that quantum computing poses no threat to symmetric algorithms.

Secure communication

Symmetric cryptography

Communicating securely with symmetric cryptography is easy. All communicating parties must share the same secret key. Plaintext is encrypted with the secret key to create ciphertext and then transmitted or stored. It can be decrypted by anyone who has the secret key.

Asymmetric cryptography

Communicating securely with asymmetric cryptography is a bit different. Anything encrypted with one key can be decrypted only by the other related key. For Alice to encrypt a message for Bob, she encrypts it with Bob’s public key. Only Bob has the corresponding key that can decrypt the message: Bob’s private key.

Hybrid cryptography

Asymmetric cryptography alleviates the problem of transmitting a key over an unsecure channel. However, it is considerably slower than symmetric cryptography. AES, for example, is approximately 1,500 times faster for decryption than RSA and 40 times faster for encryption. AES is also much faster than ECC. Key generation is also far slower with RSA or ECC than it is with symmetric algorithms, where the key is just a random number rather than a set of carefully chosen numbers with specific properties. Moreover, certain keys with RSA may be weaker than others.

Because of these factors, RSA and ECC are almost never used to encrypt large chunks of information. Instead, it is common to use hybrid cryptography, where a public key algorithm is used to encrypt a randomly-generated key that will encrypt the message with a symmetric algorithm. This randomly-generated key is called a session key, since it is generally used for one communication session and then discarded.

Key Exchange

The biggest problem with symmetric cryptography is key distribution. For Alice and Bob to communicate, they must share a secret key that no adversaries can get. However, Alice cannot send the key to Bob since it would be visible to adversaries. She cannot encrypt it because Alice and Bob do not share a key yet.

Key exchange using a trusted third party

For two parties to communicate using symmetric ciphers they need to share the same key. The ways of doing this are:

1. Share the key via some trusted mechanism outside of the network, such are reading it over the phone or sending a flash drive via FedEx.

2. Send the key using a public key algorithm.

3. Use a trusted third party.

We will first examine the use of a trusted third party. A trusted third party is a trusted system that has everyone’s key. Hence, only Alice and the trusted party [whom we will call Trent] have Alice’s secret key. Only Bob and Trent have Bob’s secret key.

The simplest way of using a trusted third party is to ask it to come up with a session key and send it to the parties that wish to communicate. For example, Alice sends a message to Trent requesting a session key to communicate with Bob. This message is encrypted with Alice’s secret key so that Trent knows the message could have only come from Alice.

Trent generates a random session key and encrypts it with Alice’s secret key. He also encrypts the same key with Bob’s secret key. Alice gets both keys and passes the one encrypted for Bob to Bob. Now Alice and Bob have a session key that was encrypted with each of their secret keys and they can communicate by encrypting messages with that session key.

This simple scheme is vulnerable to replay attacks. An eavesdropper, Eve, can record messages from Alice to Bob and replay them at a later time. Eve might not be able to decode the messages but she can confuse Bob by sending him seemingly valid encrypted messages.

The second problem is that Alice sends Trent an encrypted session key but Trent has no idea that Alice is requesting to communicate with him. While Trent authenticated Alice [simply by being able to decrypt her request] and authorized her to talk with Bob [by generating the session key], that information has not been conveyed to Bob.

Needham-Schroeder: nonces

The Needham-Schroeder protocol improves the basic key exchange protocol by adding nonces to messages. A nonce is simply a random string – a random bunch of bits. Alice sends a request to Trent, asking to talk to Bob. This time, it doesn’t have to even be encrypted. As part of the request she sends a nonce.

Trent responds with a message that contains:

• Alice’s ID
• Bob’s ID
• the nonce
• the session key
• a ticket: a message encrypted for Bob containing Alice’s ID and the same session key

This entire message from Trent is encrypted with Alice’s secret key. Alice can validate that the message is a response to her message because:

• It is encrypted for her: nobody but Alice and Trent has Alice’s secret key.
• It contains the same nonce as in her request, so it is not a replay of some earlier message, which would have had a different randomly-generated nonce.

Alice sends the ticket [the message encrypted with Bob’s key] to Bob. He can decrypt it and knows:

• The message must have been generated by Trent since only Trent and Bob know Bob’s key and and thus could construct a meaningful message encrypted with Bob’s key.
• He will be communicating with Alice because Trent placed Alice’s ID in that ticket.
• The session key since Trent placed that in the ticket as well. Alice has this too.

Bob can now communicate with Alice but he will first authenticate Alice to be sure that he’s really communicating with her. He’ll believe it’s Alice if she can prove that she has the session key. To do this, Bob creates another nonce, encrypts it with the session key, and sends it to Alice. Alice decrypts the message, subtracts one from the nonce, encrypts the result, and sends it back to Bob. She just demonstrated that she could decrypt a message using the session key and return back a known modification of the message. Needham-Schroeder is a combined authentication and key exchange protocol.

Denning-Sacco modification: timestamps to avoid key replay

One flaw in the Needham-Schroeder algorithm is when Alice sends the ticket to Bob. The ticket is encrypted with Bob’s secret key and contains Alice’s ID as well as the session key. If an attacker grabbed a communication session and managed to decrypt the session key, she can replay the transmission of the ticket to Bob. Bob won’t know that he received that same session key in the past. He will proceed to validate “Alice” by asking her to prove that she indeed knows the session key. In this case, Eve, our eavesdropper, does know it; that’s why she sent the ticket to Bob. Bob completes the authentication and thinks he is talking with Alice when in reality he is talking to Eve.

A fix for this was proposed by Denning & Sacco: add a timestamp to the ticket. When Trent creates the ticket that Alice will give to Bob, it is a message encrypted for Bob and contains Alice’s ID, the session key, and a timestamp.

When Bob receives a ticket, he checks the timestamp. If it is older than some recent time [e.g., a few seconds], Bob will simply discard the ticket, assuming that he is getting a replay attack.

Otway-Rees protocol: session IDs instead of timestamps

A problem with timestamps is that their use relies on all entities having synchronized clocks. If Bob’s clock is significantly off from Trent’s, he may falsely accept or falsely reject a ticket that Alice presents to him. Time synchronization becomes an attack vector for this protocol. If an attacker can change Bob’s concept of time, she may be able to convince Bob to accept an older ticket. To do this, she can create fake NTP [network time protocol] responses to force Bob’s clock to synchronize to a different value or, if Bob is paranoid and uses a GPS receiver to synchronize time, create fake GPS signals.

A way to avoid the replay of the ticket without using timestamps is to add a session ID to each message. The rest of the Otway-Rees protocol differs a bit from Needham-Schroeder but is conceptually very similar.

1. Alice sends a message to Bob that contains:
   • A session ID
   • Aoth of their IDs
   • A message encrypted with Alice’s secret key. This encrypted message contains Alice and Bob’s IDs as well as the session ID.
2. Bob sends Trent a request to communicate with Alice, containing:
   • Alice’s message
   • A message encrypted with Bob’s secret key that also contains the session ID.

3. Trent now knows that Alice wants to talk to Bob since the session ID is inside her encrypted message and that Bob agrees to talk to Alice since that same session ID is inside his encrypted message.

4. Trent creates a random session key encrypted for Bob and the same key encrypted for Alice and sends both of those to Bob, along with the session key.

The protocol also incorporates nonces to ensure that there is no replay attack on Trent’s response even if an attacker sends a message to Bob with a new session ID and old encrypted session keys [that were cracked by the attacker].

Kerberos

Kerberos is a trusted third party authentication, authorization, and key exchange protocol using symmetric cryptography and based closely on the Needham-Schroeder protocol with the Denning-Sacco modification [the use of timestamps].

When Alice wands to talk with Bob [they can be users and services], she first needs to ask Kerberos. If access is authorized, Kerberos will send her two messages. One is encrypted with Alice’s secret key and contains the session key for her communication with Bob. The other message is encrypted with Bob’s secret key. Alice cannot read or decode this second message. It a ticket [sometimes known as a sealed envelope]. It contains the same session key that Alice received but is encrypted for Bob. Alice will send that to Bob. When Bob decrypts it, he knows that the message must have been generated by an entity that knows its secret key: Kerberos. Now that Alice and Bob both have the session key, they can communicate securely by encrypting all traffic with that session key.

To avoid replay attacks, Kerberos places a timestamp in Alice’s response and in the ticket. For Alice to authenticate herself to Bob, she needs to prove that she was able to extract the session key from the encrypted message Kerberos sent her. She proves this by generating a new timestamp, encrypting it with the session key, and sending it to Bob. Bob now needs to prove to Alice that he can decode messages encrypted with the session key. He takes Alice’s timestamp, adds one [just to permute the value], and sends it back to Alice, encrypted with their session key.

Since your secret key is needed to decrypt every service request you make of Kerberos, you’ll end up typing your password each time you want to access a service. Storing the key in a file to cache it is not a good idea. Kerberos handles this by splitting itself into two components that run the same protocol: the authentication server [AS] and the ticket granting server [TGS]. The authentication server handles the initial user request and provides a session key to access the TGS. This session key can be cached for the user’s login session and allows the user to send requests to the TGS without re-entering a password. The TGS is the part of Kerberos that handles requests for services. It also returns two messages to the user: a different session key for the desired service and a ticket that must be provided to that service.

Diffie-Hellman key exchange

The Diffie-Hellman key exchange algorithm allows two parties to establish a common key without disclosing any information that would allow any other party to compute the same key. Each party generates a private key and a public key. Despite their name, these are not encryption keys; they are just numbers. Diffie-Hellman does not implement public key cryptography. Alice can compute a common key using her private key and Bob’s public key. Bob can compute the same common key by using his private key and Alice’s public key.

Diffie-Hellman uses the one-way function abmod c. Its one-wayness is due to our inability to compute the inverse: a discrete logarithm. Anyone may see Alice and Bob’s public keys but will be unable to compute their common key. Although Diffie-Hellman is not a public key encryption algorithm, it behaves like one in the sense that it allows us to exchange keys without having to use a trusted third party.

Key exchange using public key cryptography

With public key cryptography, there generally isn’t a need for key exchange. As long as both sides can get each other’s public keys from a trusted source, they can encrypt messages using those keys. However, we rarely use public key cryptography for large messages. It can, however, be used to transmit a session key. This use of public key cryptography to transmit a session key that will be used to apply symmetric cryptography to messages is called hybrid cryptography. For Alice to send a key to Bob:

1. Alice generates a random session key.
2. She encrypts it with Bob’s public key & sends it to Bob.
3. Bob decrypts the message using his private key and now has the session key.

Bob is the only one who has Bob’s private key to be able to decrypt that message and extract the session key. A problem with this is that anybody can do this. Charles can generate a random session key, encrypt it with Bob’s public key, and send it to Bob. For Bob to be convinced that it came from Alice, she can encrypt it with her private key [this is signing the message].

1. Alice generates a random session key.
2. She signs it by encrypting the key with her private key.
3. She encrypts the result with Bob’s public key & sends it to Bob.
4. Bob decrypts the message using his private key.
5. Bob decrypts the resulting message with Alice’s public key and gets the session key.

If anybody other than Alice created the message, the result that Bob gets by decrypting it with Alice’s public key will not result in a valid key for anyone. We can enhance the protocol by using a standalone signature [encrypted hash] so Bob can identify a valid key from a bogus one.

Forward secrecy

If an attacker steals, for example, Bob’s private key, he will be able to go through old messages and decrypt old session keys [the start of every message to Bob contained a session key encrypted with his public key]. Forward secrecy, also called perfect forward secrecy, is the use of keys and key exchange protocols where the compromise of a key does not compromise past session keys. There is no secret that one can steal that will allow the attacker to decrypt multiple past messages. Note that this is of value for communication sessions but not stored encrypted documents [such as email]. You don’t want an attacker to gain any information from a communication session even if a user’s key is compromised. However, the user needs to be able to decrypt her own documents, so they need to rely on a long-term key.

Diffie-Hellman enables forward secrecy. Alice and Bob can each generate a key pair and send their public key to each other. They can then compute a common key that nobody else will know and use that to communicate. Achieving forward secrecy requires single-use [ephemeral] keys. Next time Alice and Bob want to communicate, they will generate a new set of keys and compute a new common key. At no time do we rely on long-term keys, such as Alice’s secret key or RSA private key. Encrypting a session key with a long-term key, such as Bob’s public key, will not achieve forward secrecy. If an attacker ever finds Bob’s private key, she will be able to extract the session key.

Difie-Hellman is particularly good for for achieving forward secrecy because it is efficient to create new new key pairs on the fly. RSA or ECC keys can be used as well but key generation is far less efficient. Because of this, RSA and ECC keys tend to be used mainly as long-term keys [e.g., for authentication].

Message Integrity

One-way functions

A one-way function is one that can be computed relatively easily in one direction but there is no known way of computing the inverse function. One-way functions are crucial in a number of cryptographic algorithms, including digital signatures, Diffie-Hellman key exchange, and both RSA and elliptic curve public key cryptography. For Diffie-Hellman and public key cryptography, they ensure that someone cannot generate the corresponding private key when presented with a public key. Key exchange and asymmetric cryptography algorithms rely on a spacial form of one-way function, called a trapdoor function. This is a function whose inverse is computable if you are provided with extra information, such as a private key that corresponds to the public key that was used to generate the data.

Hash functions

A particularly useful form of a one-way function is the cryptographic hash function. This is a one-way function whose output is always a fixed number of bits for any input. Hash functions are commonly used in programming to construct hash tables, which provide O[1] lookups of keys.

Cryptographic hash functions produce far longer results than those used for hash tables. Common lengths are 224, 256, 384, or 512 bits. Good cryptographic hash functions [e.g., SHA–1, SHA–2, SHA–3] have several properties:

1. Like all hash functions, take arbitrary-length input and produce fixed-length output

2. Also like all hash functions, they are deterministic; they produce the same result each time when given identical input.

3. They exhibit pre-image resistance, or hiding. Given a hash H, it should not be feasible to find a message M where H=hash[M].

4. The output of a hash function should not give any information about any of the input. For example, changing a byte in the message should not cause any predictable change in the hash value.

5. They are collision resistant. While hash collisions can exist [the number of possible hashes is smaller than than number of possible messages; see the pigeonhole principle], it is not feasible to find any two different messages that hash to the same value. Similarly, it is not feasible to modify the plaintext without changing its resultant hash.

6. They should be relatively efficient to compute. We would like to use hash functions as message integrity checks and generate them for each message without incurring significant overhead.

The cryptographic hash function is the basis for message authentication codes and digital signatures.

Because of these properties, we have extremely high assurance that a message would no longer hash to the same value if it is modified in any way. The holy grail for an attacker is to be able to construct a message that hashes to the same value as another message. That would allow the attacker to substitute a new message for some original one [for example, redirecting a money transfer]. Searching for a collision with a pre-image [known message] is much harder than searching for any two messages that produce the same hash. The birthday paradox tells us that the search for a collision of any two messages is approximately the square root of the complexity of searching for a collision on a specific message. This means that the strength of a hash function for a brute-force collision attack is approximately half the number of bits of the hash. A 256-bit hash function has a strength of approximately 128 bits.

Popular hash functions include SHA–1 [160 bits], SHA–2 [commonly 256 and 512 bits], and SHA–3 [256 and 512 bits].

Message Integrity and Hash Pointers

A cryptographic hash serves as a checksum for a message. If a message has been modified, it will yield a different hash. By associating a hash with a message, we have a basis for managing the integrity of that message: being able to detect if the message gets changed.

Tamper-resistant linked-lists: blockchains

One way of associating a hash with a message is via the use of hash pointers. Pointers are used in data structures to allow one data element to refer to another. In processes, a pointer is a memory location. In distributed systems, a pointer may be an IP address and object identifier. A hash pointer is a tuple that contains a traditional pointer along with the hash of the data element that is being pointed to. It allows us to validate that the information being pointed to has not been modified.

The same structures that use pointers can be modified to use hash pointers and create tamper-evident structures. For example, a linked list can be constructed with each element containing a hash pointer to the next element instead of a pointer.

PayPal

Mobile device security

What makes mobile devices unique?

In many ways, mobile devices should not be different from laptops or other computer systems. They run operating systems that are derived from those those systems, run multiple apps, and connect to the network. There are differences, however, that make them more attractive targets to attackers.

Users

Several user factors make phones different from most computing devices:

• Mobile users often do not think of their phones as real computers. They may not have the same level of paranoia that malware may get in or their activities may be monitored.

• Users tend to install a lot more apps on their phones than they do on their computers. These apps are more likely to be from unknown software vendors than those installed on computers.

• Social engineering may work more easily on phones. People are often in distracted environments when using their phones and may not pay attention to realize they are experiencing a phishing attack.

• Phones are small. Users may be less likely to notice some security indicators, such as an EV certificate indicator. It is also easier to lose the phone … or have it stolen.

• A lot of phones are protected with bad PINs. Four-digit PINs still dominate and, as with passwords, people tend to pick bad ones – or at least common ones. In fact, four PINs [1234, 1111, 0000, 1212, 7777] account for over 20% of PINs chosen by users.

• While phones have safeguards to protect resources that apps can access users may grant app permission requests without thinking: they will just click through during installation to get the app up and running.

Interfaces

Phones have many sensors built into them: GSM, Wi-Fi, Bluetooth, and NFC radios as well as a GPS, microphone, camera. 6-axis gyroscope and accelerometer, and even barometer. These sensors can enable attackers to monitor the world around you: identify where you are and whether you are moving. They can record conversations and even capture video. The sensors are so sensitive that it has been demonstrated that a phone on a desk next to a keyboard can pick up vibrations from a user typing on the neighboring keyboard. This led to a word recovery rate of 80%.

Apps

There are a lot of mobile apps. Currently, there are about 2.6 million Android apps and 2.2 million iOS apps. Most of these apps are written by unknown, and hence untrusted, parties. We would be be wary of downloading many of these on our PCs but think nothing of doing so on our phones. We place our trust in several areas:

• The testing & approval process by Google [automated] and Apple [automated + manual]
• The ability of the operating system to sandbox an application
• The operating system’s requirement of users granting permissions to access certain resources.

This trust may be misplaced as the approval process is far from foolproof. Overtly misadvertised or malicious apps can be detected but it is impossible to discern what a program will do in the future. Sandboxes have been broken in the past and users may be too happy to grant permissions to apps. Moreover, apps often ask for more permissions than they use. For example, a security researcher surveyed flashlight apps available for Android and discovered that, of the 937 apps surveyed, the majority requested an average of 25 permissions per app.

Most apps do not get security updates. There is little economic incentive for a developer to support existing apps, particularly if newer ones have been deployed.

Platform

Mobile phones are comparable to desktop systems in complexity. In some cases, they may even be more complex. This points to the fact that, like all large systems, they will have bugs and some of these bugs will be security sensitive. For instance, in late March, 2017, Apple released an upgrade for iOS, stating that they fixed over 80 security flaws. This is almost 10 years after the release of the iPhone. You can be certain there are many more flaws lurking in the system and more will be added as new features are introduced.

Because of bugs in the system, malicious apps may be able to get root privileges. If they do, they can install rootkits, enabling long-term control while concealing their presence. A lot of malicious iOS apps, for instance, gained root privileges by exploiting heap overflow vulnerabilities.

Unlike desktop systems and laptops, phones enforce a single user environment. Although PCs are usually used as single-user systems, they support multiple user accounts and run a general-purpose timesharing operating system. Mobile devices are more carefully tuned to the single-user environment.

Threats

Mobile devices are are threats to personal privacy as well as at risk of traditional security violations. Personal privacy threats include identifying users and user location, accessing the camera and microphone, and leaking personal data from the phone over the network. Additional threats include traditional phishing attacks, malware installation, malicious Android intents [messages to other apps or services], and overly-broad access to system resources and sensors.

Android security

Android was conceived as an operating system for network-connected smart mobile devices. The company was acquired by Google in 2005 and the engineering effort shifted to developing it as a Linux-based operating system platform that would be provided for free to third-party phone manufacturers. Google would make money from services and apps.

Applications on Android had specific needs, some of which have not been priorities in the design of desktop operating systems. These include:

IntegrityThe platform should ensure that the app is not modified between the time of its creation and its installation by users.IsolationEach app needs private data and app components as well as the private data need to be protected from other apps.SharingApps may need access to shared storage and devices, including the network. This includes the on-device file system, external storage, communication interfaces, and the various sensors available on phones and other smart devices.Inter-app servicesAn app needs to be able to send messages to other apps to interact with services – but only when it is permitted to do so. This affects the design of apps. Desktop apps have a single launching point and generally run as single, monolithic process. Android apps, on the other hand, contain multiple independent components that include activities [user-facing parts], services [background components], and content providers [data access components].PortabilityThe previous needs are general and apply to other mobile platforms, such as iOS. Since Android was designed as an operating system for a variety of hardware, apps should be able to run on different hardware architectures. A decision was made that apps should be distributed in a portable, architecture-neutral format. Apps run under the ART [Android Runtime] virtual machine, which is a variant of of the Java virtual machine [JVM]. The original intention in Android was that apps would be written only in Java but it soon became clear that support for native [C and C++] code was needed and Google introduced the Native Development Kit to support this.

App package

An Android app is distributed as a package in an APK format, which is a single zip-format compressed file that contains several components:

ActivityCode for the user-visible component - the interface. The code component includes compiled code and needed resource files, such as strings, images, UI layouts, and data.ServiceCode background component - services that the app may offer to other programsContent providerA database for whatever persistent data the app needs to store. It implements APIs to provide access to structured data. For instance, user contacts will be stored in a content provider.Broadcast receiverMailbox for received messagesPackage manifest [META-INF]This contains items needed to validate the origin of the package and that it has not been tampered: - Signed list of hashes - Application creator’s certificateApplication manifestThis enumerates what makes up the application and includes: - Name [e.g., com.example.myapp] - Components list - Device requirements - Intents: interfaces the app supports to activate services & activities - Permissions: access to services this app requires [e.g., android.permission.SEND_SMS] - Permissions other apps need to access services this app provides

App Integrity

All Android apps must be signed by the developer. This allows both developers and users to know that the application will be installed without modifications on their device. On installation, the Android Package Manager verifies the signature.

Applications do not have to be signed by any central authorities. Developers can use a self-signed certificate and Android does not perform verification of CAs [certification authorities] in the certificate.

Prior to distribution, the contents of the APK package are hashed and signed by the developer and then the signature along with the developer’s certificate is inserted into the APK.

The app package is verified prior to installation by creating a hash and validating it against the signature in the package. If the app is distributed through the Google Play store, Google performs the same checks.

For additional protection, Google optionally supports a feature called Google Play Protect. This validates the app before it is downloaded and checks the user’s device for potential malware. It warns the user about malware or any apps that violate Google’s Unwanted Software Policy, such as apps that hide or misrepresent information.

The Android app sandbox

Android relies on process sandboxing for most of its security. Android is based on Linux, which is a multi-user operating system. Under Linux, each user has a distinct user ID and all apps run by that user run with the privileges of the user [ignoring setUID apps]. This allows any one app full access to all user data.

User IDs

Android supports only a single user and uses Linux user IDs for isolating app privileges. Under Android, each app normally runs under a different user ID. Hence, apps are isolated and can only access their resources. Access requests to other objects involve messages that pass through a gatekeeper, which validates access requests.

Core Android services also run in a similar manner, with each service running under its own unique user ID. For example:

user idservice

1001	Telephony
1002	Bluetooth
1003	Graphics
1004	Input devices
1005	Audio

Related apps may share the same Linux user ID if a sharedUserID attribute is set to the same domain for two or more applications as long as those apps are also signed by the same certificate. This would allow these related apps to share files and they can be configured to even share the same Dalvik virtual machine.

File permissions

Two mechanisms are used to enforce file access permissions:

1. Linux file permissions These provide discretionary access control, allowing the owner [and root] to change permissions to allow others access to the files. With this mechanism, an app can decide to share a data file.

2. SELinux mandatory access control Certain data and cache directories in Android are protected with the SELinux [Security-Enhanced Linux] mandatory access control [MAC] kernel extension. This ensures that even the owner cannot change access permissions for the files.

Internal storage provides a per-app private directory for files used by each application. External storage [e.g., attached microSD cards or USB devices] is shared among all apps and, of course, may be moved to other computers.

Intents

Android apps communicate with system services, between app components, and with other apps via intents. Intents are messaging objects. An intent is a message that contains: - requested action - data being sent to the action - the component and app that should handle the intent

Intents are declarations of app capabilities and the messaging format used to communicate with an app. They identify app components and how those components are started [e.g., foreground or background and what the entry points are].

Intents allow an app to:

• Start a service [background task]
• Start an activity [start a user-facing foreground task, such as a camera or phone]
• Deliver notifications to one or more apps [broadcasts]

An app lists its available intents in its app manifest and these intents are registered when the app is installed. The intents form the list of services that the app exposes to other applications. If several apps register the same intent, the user selects which app should be launched. For example, you may have multiple browsers installed and are asked to pick the one that should be associated with implicit intents to display a URL.

Intents may explicit or implicit. With explicit intents, the app identifies the target component in the intent. That is, the intent is sent to a specific, named app. With implicit intents, the app asks Android to find a component based on the type of data being sent. For example, sending a URL to display a web page can be an explicit intent that will cause Android to open the default web browser.

Intents are used to invoke system services as well as services available on any installed apps. Common examples of intents are: add a calendar event, set an alarm, take a photo & return it, view a contact, add a contact, show a location on a map, retrieve a file, or initiate a phone call.

Intents pass through & are validated by Android’s gatekeeper.

Permissions

An app manifest also contains permissions, which can specify which apps or services are allowed access to the app’s services and whether a user needs to be prompted to grant permission. Permissions determine whether one app is allowed to access another app’s component.

Apps need permissions to access any services, which include:

• System resources: logs, battery levels, …
• System interfaces: Internet, Bluetooth, send SMS, send email, …
• Sensitive data: SMS messages, contacts, email, …
• Any app-defined services

Every service, whether a normal app or a system service is assigned a protection level that determines who may be able to access it:

PermissionType

Normal	this is the default; there is no danger to the users or system if this service is accessed
Dangerous	Access can compromise the system or privacy. The user has to approve access during installation or runtime
Signature	Access granted only if the app signed by the same developer & contains the same certificate. This allows related apps to share services [e.g., Microsoft Office apps]
SignatureOrSystem	Similar to signature_ but access will be granted if a system application is requesting it

The application manifest file defines the type of permission and the service that is associated with permission name.

Permissions are managed in two forms:

Permission text stringsThese are enforced by Android middleware. Sensitive resources such as the phone are only accessible via APIs and access is mediated through these APIs.Linux group IDsGroup permissions are enforced by Linux file access checks. For efficiency, networking and file access operations do not go through APIs but directly to Linux. This includes access to Bluetooth, Wi-Fi, and external storage. To be able to access resources, the app needs to be a member of the group that corresponds to the resource. Android dynamically adds user IDs to various groups based on what permissions are granted to them.

Other protections

The Linux operating system provides per-process memory isolation and address space layout randomization [ASLR]. Linux also uses no-execute [NX] protection on stack and heap memory pages if the processor supports it

The Java compiler provides provides stack canaries, and its memory management libraries provide some heap overflow protections [checks of backward & forward pointers in dynamically allocated structures].

Android supports whole disk encryption so that if a device is stolen, an attacker will not be able to easily recover file contents even with raw access to the flash file system.

Unlike iOS, Android supports the concurrent execution of multiple apps. It is up to the developer to think about being frugal with battery life. Apps store state their state in persistent memory so they can be stopped and restarted at any time. This ability to stop an app also helps with DoS attacks as the app is not accepting requests or using system resources.

Security concerns

An app can probe whether another app has specific permissions by specifying a permission with an intent method call to that app. This can help an attacker identify a target app. Receivers need to be able to handle malicious intents, even for actions they do not expect to handle and for data that might not make sense for the action.

Apps may also exploit permissions re-delegation. An app, not having a certain permission, may be able gain those privileges by communicating through another app. If a public component does not explicitly have an access permission listed in its manifest definition, Android permits any app to access it. For example, the Power Control Widget [a default Android widget] allows third-party apps to change protected system settings without requesting permissions to control those settings. This is done by presenting the user with a pop-up interface to control power-related settings. A malicious app can send a fake intent to the Power Control Widget while simulating the pressure of the widget button to switch power-related settings. It is effectively simulating a user’s actions on the screen.

By using external storage, apps can exercise permissions avoidance. By default, all apps have access to external storage. Many apps store data in external storage without specifying any protection, allowing other apps to access that data.

Another way permissions avoidance is used is that Android intents allow opening some system apps without requiring permission to do so. These apps include the camera, SMS, contact list, and browser. For instance, opening a browser via an intent can be dangerous since it enables data transmission, receiving remote commands, and even downloading files without user intervention.

iOS security

App signing

iOS requires mandatory code signing. Unlike Android, which accepts self-signed certificates, the app package must be signed using an Apple Developer certificate and apps are only available for This does not ensure trustworthiness of an app but identifies the registered developer and ensures that the app has not been modified after it has been signed.

Runtime protection

Apple’s iOS provides runtime protection via OS-level sandboxing. System resources and the kernel are shielded from user apps. The sandbox also limits which system calls an app can call Except through kernel exploits, an app cannot leave its sandbox.

The app sandbox restricts the ability of one app to access another app’s data and resources. Each app has its own sandbox directory. The OS enforces the sandbox and permits access only to files within that directory, as well as restricted access to to system preferences, the network, and other resources.

Inter-app communication can take place only through iOS APIs. Code generation by an app is prevented because data memory pages cannot be made executable and executable memory pages are not writable by user processes.

Data protection

All file contents are encrypted with a unique 256-bit AES per-file key, which is generated when the file is created.

This per-file key is encrypted with a class key and is stored along with the file’s metadata, which is the part of the file system that describes attributes of the file, such as size, modification time, and access permissions.

The class key is generated from a hardware key in the device and the user’s passcode. Unless the passcode is entered, the class key cannot be created and the file key cannot be decrypted.

The file system’s metadata is also encrypted. A file system key is used for this, which is derived directly from the hardware key, which is generated when iOS is installed. Keys are stored in Apple’s Secure Enclave, a separate processor and isolated memory that cannot be accessed directly by the main processor. Encrypting metadata encrypts the entire structure of the file system. Someone who rips out the flash memory from an iOS device and examines it will not be able to see neither file contents [they are encrypted with per-file keys] nor information about those files [the metadata is encrypted with a file system key].

A hardware AES engine encrypts and decrypts the file as it is written/read on flash memory so file encryption is done transparently and efficiently.

The iOS kernel partition is mounted read-only, so even if an app manages to break out of its sandbox due to some vulnerability and gain root access, it will still not have permission to modify the kernel.

Communication

The iOS sandbox restricts apps from accessing files stored by other apps or making changes to device settings. Each app is given a unique home directory for its files. System files and resources are shielded from the user’s apps

Unlike Android, where each app is assigned a unique user ID, apps under iOS run as a non-privileged user “mobile.”

The iOS framework grants entitlements to apps. These are digitally-signed key-value pairs that are granted to an app to allow access to specific services. It is essentially a capability. If you have an entitlement, then you can access a service.

Kernel protection

In addition to the sandbox, iOS also uses address space layout randomization [ASLR] and memory execute protection for stack and heap pages via ARM’s Execute Never [XN] memory page flag.

Masque attacks

While Apple normally expects users to install apps only from its App Store, users need to be able to deploy pre-production apps to friendly parties for testing and enterprises may need to deploy in-house apps to their employees. Apple supports a Developer Enterprise Program to create and distribute such in-house apps. This mechanism has been used to replace existing apps with private versions. The vulnerability has been patched.

iOS has been hit several times with masque attacks. While there have been various forms of these, the basic attack is to get users to install malicious apps that have been created with the same bundle identifier as some exiting legitimate app. This malicious app replaces the legitimate app and masquerades as that app. Since Apple will not host an app with a duplicate bundle identifier, the installation of these apps has to bypass the App Store. Enterprise provisioning is used to get users to install this. which typically requires the user going to a URL that redirects the user to an XML manifest file hosted on a server. The ability to launch this attack is somewhat limited as the user will generally need to have an enterprise certificate installed to make the installation seamless.

Web apps

Both iOS and Android have full web browsers that can be used to access web applications. They also permit web apps to appear as a regular app icon. The risks here are the same as those for web browsers in general: loading untrusted content and leaking cookies and URLs to foreign apps.

Mobile-focused web-based attacks can take advantage of the sensors on phones. The HTML5 Geolocation API allows JavaScript to find your location. A Use Current Location permission dialog appears, so the attacker has to hope the user will approve but there the attacker can provide incentives via a Trojan horse approach: provide a service that may legitimately need your location.

Recently, a proof of concept web attack showed how JavaScript could access the phone’s accelerometers to detect movements of the phone as a user enters a PIN. The team that implemented this achieved a 100% success rate of recognizing a four-digit PIN within five attempts of a user entering it. Apple patched this specific vulnerability but there may be more undiscovered ones.

Hardware support for security

ARM TrustZone worlds

All Android and iOS phones currently use ARM processors. ARM provides a dedicated security module, called TrustZone, that coexists with the normal processor. The hardware is separated into two “worlds”: secure [trusted] and non-secure [non-trusted] worlds. Any software resides in only one of these two worlds and the processor executes in only one world at a time.

Each of these worlds has its own operating system and applications. Android systems run an operating system called Trusty TEE in the secure world and, of course, Linux in the untrusted world.

Logically, you can think of the two worlds as two distinct processors, each running their own operating system with their own data and their own memory. Non-secure applications cannot access any own memory or registers of secure resources directly. The only way they can communicate is through a messaging API.

In practice, the hardware creates two virtual cores for each CPU core, managing separate registers and all processing state in each world.

The phone’s operating system and all applications reside in the non-trusted world. Secure components, such as cryptographic keys, signature services, encryption services, and payment services live in the trusted world. Even the operating system kernel does not have access to any of the code or data in the trusted world. Hence, even if an app manages a privilege escalation attack and gains root access, it will be unable to access certain security-critical data.

Applications for the trusted world include key management, secure boot, digital rights management, secure payment processing, mobile payments, and biometric authentication.

Apple Secure Enclave

Apple uses modified ARM processors for iPhones and iPads. In 2013, they announced Secure Enclave for their processors. The details are confidential but it appears to be similar in function to ARM’s TrustZone but designed as a physically separate coprocessor. As with TrustZone, the Secure Enclave coprocessor runs its own operating system [a modified L4 microkernel in this case].

The processor has its own secure bootloader and custom software update mechanism. It uses encrypted memory so that anything outside the Secure Enclave cannot access its data. It provides:

• All cryptographic operations for data protection & key management.
• Random number generation.
• Secure key store, including Touch ID [fingerprint] and the Face ID neural network.
• Data storage for payment processing.

The Secure Enclave maintains the confidentiality and integrity of data even if the iOS kernel has been compromised.

Steganography and Watermarking

Cryptography’s goal is to hide the contents of a message. Steganography’s goal is to hide the very existence of the message. Classic techniques included the use of invisible ink, writing a message on one’s head and allowing the hair to cover it, microdots, and carefully-clipped newspaper articles that together communicate the message.

A null cipher is one where the actual message is hidden among irrelevant data. For example, the message may comprise the first letter of each word [or each sentence, or every second letter, etc.]. Chaffing and winnowing entails the transmission of a bunch of messages, of which only certain ones are legitimate. Each message is signed with a key known only to trusted parties [e.g., a MAC]. Intruders can see the messages but can’t validate the signatures to distinguish the valid messages from the bogus ones.

Messages can be embedded into images. There are a couple of ways of hiding a message in an image:

1. A straightforward method to hide a message in an image is to use low-order bits of an image, where the user is unlikely to notice slight changes in color. An image is a collection of RGB pixels. You can mess around with the least-significant bits and nobody will notice changes in the image, so you can just encode the entire message by spreading the bits of the message among the least-significant bits of the image.

2. You can do a similar thing but apply a frequency domain transformation, like JPEG compression does, by using a Discrete Cosine Transform [DCT]. The frequency domain maps the image as a collection ranging from high-frequency areas [e.g., “noisy” parts such as leaves, grass, and edges of things] through low-frequency areas [e.g., a clear blue sky]. Changes to high frequency areas will mostly be unnoticed by humans: that’s why jpeg compression works. It also means that you can add your message into those areas and then transform it back to the spatial domain. Now your message is spread throughout the higher-frequency parts of the image and can be extracted if you do the DCT again and know where to look for the message.

Many laser printers embed a serial number and date simply by printing very faint color splotches.

Steganography is closely related to watermarking. and the terms “steganography” and “watermarking” are often used interchangeably.

The primary goal of watermarking is to create an indelible imprint on a message such that an intruder cannot remove or replace the message. It is often used to assert ownership, authenticity, or encode DRM rules. The message may be, but does not have to be, invisible.

The goal of steganography is to allow primarily one-to-one communication while hiding the existence of a message. An intruder – someone who does not know what to look for – cannot even detect the message in the data.

References

Injection

SQL Injection, The Open Web Application Security Project, April 10, 2016.

SQL Injection, Acunetix.

Simson Garfinkel & Gene Spafford, Section 11.5, Protecting Yourself, Practical UNIX & Internet Security, Second Edition, April 1996. Discusses shell attacks.

Directory traversal attack, Wikipedia.

Why does Directory traversal attack %C0%AF work?, Information Security Stack Exchange, September 9, 2016

Tom Rodriquez, What are unicode vulnerabilities on Internet Information Server [IIS]?, SANS.

The Unicode Consortium.

IDN homograph attack, Wikipedia.

Time of check to time of use, Wikipedia.

Michael Cobb, How to mitigate the risk of a TOCTTOU attack, TeachTarget, August 2011.

Ernst & Yount LLP Security & Technology Solutions, Using Attack Surface Area And Relative Attack Surface Quotient To Identify Attackability. Customer Information Paper.

Michael Howard, Back to the Future: Attack Surface Analysis and Reduction, Microsoft Secure Blog, February 14, 2011.

Olivier Sessink, Jailkit, November 18, 2015.

Confinement

Evan Sarmiento, Chapter 4. The Jail Subsystem, FreeBSD Architecture Handbook, The FreeBSD Documentation Project. 2001, Last modified: 2016–10–29.

Matteo Riondato, Chapter 14. System Administration: Jails, FreeBSD Architecture Handbook, The FreeBSD Documentation Project. 2001, Last modified: 2016–10–29.

Chapter 1. Introduction to Control Groups, Red Hat Enterprise Linux 6.8 Resource Management Guide.

José Manuel Ortega, Everything you need to know about Containers Security, OSDEM 2018 presentation.

Johan De Gelas, Hardware Virtualization: the Nuts and Bolts, AnandTech article, March 17 2008.

Which location contains configuration information for sendmail?

Which one of the following tools can be programmed to examine TCP headers to FIN The SYN flag?

What type of attacks use every possible letter number and character found on a keyboard when cracking a password?

Which format can most packet analyzer tools read?