Securing coding is the practice of developing computer software in a way that guards against the accidental introduction of security vulnerabilities. Defects, bugs and logic flaws are consistently the primary cause of commonly exploited software vulnerabilities. Through the analysis of thousands of reported vulnerabilities, security professionals have discovered that most vulnerabilities stem from a relatively small number of common software programming errors. By identifying the insecure coding practices that lead to these errors and educating developers on secure alternatives, organizations can take proactive steps to help significantly reduce or eliminate vulnerabilities in software before deployment.
Techniques
Various techniques have been used to detect or prevent buffer overflows, with various tradeoffs. The most reliable way to avoid or prevent buffer overflows is to use automatic protection at the language level. This sort of protection, however, cannot be applied to legacy code, and often technical, business, or cultural constraints call for a vulnerable language. The following sections describe the choices and implementations available.
Choice of programming language – The choice of programming language can have a profound effect on the occurrence of buffer overflows. As of 2008, among the most popular languages are C and its derivative, C++, with a vast body of software having been written in these languages. C provides no built-in protection against accessing or overwriting data in any part of memory; more specifically, it does not check that data written to a buffer is within the boundaries of that buffer. The standard C++ libraries provide many ways of safely buffering data, and C++’s Standard Template Library (STL) provides containers that can optionally perform bounds checking if the programmer explicitly calls for checks while accessing data. For example, a vector’s member function at() performs a bounds check and throws an out_of_range exception if the bounds check fails. However, C++ behaves just like C if the bounds check is not explicitly called. Techniques to avoid buffer overflows also exist for C.
Many other programming languages provide runtime checking and in some cases even compile-time checking which might send a warning or raise an exception when C or C++ would overwrite data and continue to execute further instructions until erroneous results are obtained which might or might not cause the program to crash. Examples of such languages include Ada, Eiffel, Lisp, Modula-2, Smalltalk, OCaml and such C-derivatives as Cyclone, Rust and D. The Java and .NET Framework bytecode environments also require bounds checking on all arrays. Nearly every interpreted language will protect against buffer overflows, signaling a well-defined error condition. Often where a language provides enough type information to do bounds checking an option is provided to enable or disable it. Static code analysis can remove many dynamic bound and type checks, but poor implementations and awkward cases can significantly decrease performance. Software engineers must carefully consider the tradeoffs of safety versus performance costs when deciding which language and compiler setting to use.
Use of safe libraries – The problem of buffer overflows is common in the C and C++ languages because they expose low level representational details of buffers as containers for data types. Buffer overflows must thus be avoided by maintaining a high degree of correctness in code which performs buffer management. It has also long been recommended to avoid standard library functions which are not bounds checked, such as gets, scanf and strcpy. The Morris worm exploited a gets call in fingerd.
Well-written and tested abstract data type libraries which centralize and automatically perform buffer management, including bounds checking, can reduce the occurrence and impact of buffer overflows. The two main building-block data types in these languages in which buffer overflows commonly occur are strings and arrays; thus, libraries preventing buffer overflows in these data types can provide the vast majority of the necessary coverage. Still, failure to use these safe libraries correctly can result in buffer overflows and other vulnerabilities; and naturally, any bug in the library itself is a potential vulnerability. “Safe” library implementations include “The Better String Library”, Vstr and Erwin. The OpenBSD operating system’s C library provides the strlcpy and strlcat functions, but these are more limited than full safe library implementations.
In September 2007, Technical Report 24731, prepared by the C standards committee, was published; it specifies a set of functions which are based on the standard C library’s string and I/O functions, with additional buffer-size parameters. However, the efficacy of these functions for the purpose of reducing buffer overflows is disputable; it requires programmer intervention on a per function call basis that is equivalent to intervention that could make the analogous older standard library functions buffer overflow safe.
Buffer overflow protection – Buffer overflow protection is used to detect the most common buffer overflows by checking that the stack has not been altered when a function returns. If it has been altered, the program exits with a segmentation fault. Three such systems are Libsafe, and the StackGuard and ProPolice gcc patches.
Microsoft’s implementation of Data Execution Prevention (DEP) mode explicitly protects the pointer to the Structured Exception Handler (SEH) from being overwritten.
Stronger stack protection is possible by splitting the stack in two: one for data and one for function returns. This split is present in the Forth language, though it was not a security-based design decision. Regardless, this is not a complete solution to buffer overflows, as sensitive data other than the return address may still be overwritten.
Pointer protection – Buffer overflows work by manipulating pointers (including stored addresses). PointGuard was proposed as a compiler-extension to prevent attackers from being able to reliably manipulate pointers and addresses. The approach works by having the compiler add code to automatically XOR-encode pointers before and after they are used. Because the attacker (theoretically) does not know what value will be used to encode/decode the pointer, he cannot predict what it will point to if he overwrites it with a new value. PointGuard was never released, but Microsoft implemented a similar approach beginning in Windows XP SP2 and Windows Server 2003 SP1. Rather than implement pointer protection as an automatic feature, Microsoft added an API routine that can be called at the discretion of the programmer. This allows for better performance (because it is not used all of the time), but places the burden on the programmer to know when it is necessary.
Because XOR is linear, an attacker may be able to manipulate an encoded pointer by overwriting only the lower bytes of an address. This can allow an attack to succeed if the attacker is able to attempt the exploit multiple times or is able to complete an attack by causing a pointer to point to one of several locations (such as any location within a NOP sled). Microsoft added a random rotation to their encoding scheme to address this weakness to partial overwrites.
Executable space protection – Executable space protection is an approach to buffer overflow protection which prevents execution of code on the stack or the heap. An attacker may use buffer overflows to insert arbitrary code into the memory of a program, but with executable space protection, any attempt to execute that code will cause an exception.
Some CPUs support a feature called NX (“No eXecute”) or XD (“eXecute Disabled”) bit, which in conjunction with software, can be used to mark pages of data (such as those containing the stack and the heap) as readable and writeable but not executable.
Some Unix operating systems (e.g. OpenBSD, OS X) ship with executable space protection (e.g. W^X). Some optional packages includes PaX, Exec Shield and Openwall. Newer variants of Microsoft Windows also support executable space protection, called Data Execution Prevention. Proprietary add-ons includes BufferShield and StackDefender
Executable space protection does not generally protect against return-to-libc attacks, or any other attack which does not rely on the execution of the attackers code. However, on 64-bit systems using ASLR, as described below, executable space protection makes it far more difficult to execute such attacks.
Address space layout randomization – Address space layout randomization (ASLR) is a computer security feature which involves arranging the positions of key data areas, usually including the base of the executable and position of libraries, heap, and stack, randomly in a process’ address space.
Randomization of the virtual memory addresses at which functions and variables can be found can make exploitation of a buffer overflow more difficult, but not impossible. It also forces the attacker to tailor the exploitation attempt to the individual system, which foils the attempts of internet worms. A similar but less effective method is to rebase processes and libraries in the virtual address space.
Deep packet inspection – The use of deep packet inspection (DPI) can detect, at the network perimeter, very basic remote attempts to exploit buffer overflows by use of attack signatures and heuristics. These are able to block packets which have the signature of a known attack, or if a long series of No-Operation instructions (known as a nop-sled) is detected, these were once used when the location of the exploit’s payload is slightly variable. Packet scanning is not an effective method since it can only prevent known attacks and there are many ways that a ‘nop-sled’ can be encoded. Shellcode used by attackers can be made alphanumeric, metamorphic, or self-modifying to evade detection by heuristic packet scanners and intrusion detection systems.
Best Practices
These best practices or principles, are to be followed for secured software development
- Minimize attack surface area – Every feature that is added to an application adds a certain amount of risk to the overall application. The aim for secure development is to reduce the overall risk by reducing the attack surface area. For example, a web application implements online help with a search function. The search function may be vulnerable to SQL injection attacks. If the help feature was limited to authorized users, the attack likelihood is reduced. If the help feature’s search function was gated through centralized data validation routines, the ability to perform SQL injection is dramatically reduced. However, if the help feature was re-written to eliminate the search function (through better user interface, for example), this almost eliminates the attack surface area, even if the help feature was available to the Internet at large.
- Establish secure defaults – There are many ways to deliver an “out of the box” experience for users. However, by default, the experience should be secure, and it should be up to the user to reduce their security – if they are allowed. For example, by default, password aging and complexity should be enabled. Users might be allowed to turn these two features off to simplify their use of the application and increase their risk.
- Principle of Least privilege – The principle of least privilege recommends that accounts have the least amount of privilege required to perform their business processes. This encompasses user rights, resource permissions such as CPU limits, memory, network, and file system permissions. For example, if a middleware server only requires access to the network, read access to a database table, and the ability to write to a log, this describes all the permissions that should be granted. Under no circumstances should the middleware be granted administrative privileges.
- Principle of Defense in depth – The principle of defense in depth suggests that where one control would be reasonable, more controls that approach risks in different fashions are better. Controls, when used in depth, can make severe vulnerabilities extraordinarily difficult to exploit and thus unlikely to occur. With secure coding, this may take the form of tier-based validation, centralized auditing controls, and requiring users to be logged on all pages. For example, a flawed administrative interface is unlikely to be vulnerable to anonymous attack if it correctly gates access to production management networks, checks for administrative user authorization, and logs all access.
- Fail securely – Applications regularly fail to process transactions for many reasons. How they fail can determine if an application is secure or not.
- Don’t trust services – Many organizations utilize the processing capabilities of third party partners, who more than likely have differing security policies and posture than you. It is unlikely that you can influence or control any external third party, whether they are home users or major suppliers or partners. Therefore, implicit trust of externally run systems is not warranted. All external systems should be treated in a similar fashion. For example, a loyalty program provider provides data that is used by Internet Banking, providing the number of reward points and a small list of potential redemption items. However, the data should be checked to ensure that it is safe to display to end users, and that the reward points are a positive number, and not improbably large.
- Separation of duties – A key fraud control is separation of duties. For example, someone who requests a computer cannot also sign for it, nor should they directly receive the computer. This prevents the user from requesting many computers, and claiming they never arrived. Certain roles have different levels of trust than normal users. In particular, administrators are different to normal users. In general, administrators should not be users of the application.
- Avoid security by obscurity – Security through obscurity is a weak security control, and nearly always fails when it is the only control. This is not to say that keeping secrets is a bad idea, it simply means that the security of key systems should not be reliant upon keeping details hidden. For example, the security of an application should not rely upon knowledge of the source code being kept secret. The security should rely upon many other factors, including reasonable password policies, defense in depth, business transaction limits, solid network architecture, and fraud and audit controls.
- Keep security simple – Attack surface area and simplicity go hand in hand. Certain software engineering fads prefer overly complex approaches to what would otherwise be relatively straightforward and simple code. Developers should avoid the use of double negatives and complex architectures when a simpler approach would be faster and simpler.
- Fix security issues correctly – Once a security issue has been identified, it is important to develop a test for it, and to understand the root cause of the issue. When design patterns are used, it is likely that the security issue is widespread amongst all code bases, so developing the right fix without introducing regressions is essential. For example, a user has found that they can see another user’s balance by adjusting their cookie. The fix seems to be relatively straightforward, but as the cookie handling code is shared among all applications, a change to just one application will trickle through to all other applications. The fix must therefore be tested on all affected applications.