–Randy Meyers, Chairman of ANSI C
“For years we have relied upon the CERT/CC to publish advisories documenting an endless stream of security problems. Now CERT has embodied the advice of leading technical experts to give programmers and managers the practical guidance needed to avoid those problems in new applications and to help secure legacy systems. Well done!”
–Dr. Thomas Plum, founder of Plum Hall, Inc.
“Connectivity has sharply increased the need for secure, hacker-safe applications. By combining this CERT standard with other safety guidelines, customers gain all-round protection and approach the goal of zero-defect software.”
–Chris Tapp, Field Applications Engineer, LDRA Ltd.
“I’ve found this standard to be an indispensable collection of expert information on exactly how modern software systems fail in practice. It is the perfect place to start for establishing internal secure coding guidelines. You won’t find this information elsewhere, and, when it comes to software security, what you don’t know is often exactly what hurts you.”
–John McDonald, coauthor of The Art of Software Security Assessment
Software security has major implications for the operations and assets of organizations, as well as for the welfare of individuals. To create secure software, developers must know where the dangers lie. Secure programming in C can be more difficult than even many experienced programmers believe.
This book is an essential desktop reference documenting the first official release of The CERT® C Secure Coding Standard . The standard itemizes those coding errors that are the root causes of software vulnerabilities in C and prioritizes them by severity, likelihood of exploitation, and remediation costs. Each guideline provides examples of insecure code as well as secure, alternative implementations. If uniformly applied, these guidelines will eliminate the critical coding errors that lead to buffer overflows, format string vulnerabilities, integer overflow, and other common software vulnerabilities.
About the Author
Table of Contents
Preface xviiAcknowledgments xxxiAbout the Author xxxiii Chapter 1: Using This Standard 1 System Qualities 1Automatically Generated Code 2Compliance 3 Chapter 2: Preprocessor (PRE) 5 PRE00-C. Prefer inline or static functions to function-like macros 6PRE01-C. Use parentheses within macros around parameter names 11PRE02-C. Macro replacement lists should be parenthesized 13PRE03-C. Prefer type definitions to defines for encoding types 15PRE04-C. Do not reuse a standard header file name 16PRE05-C. Understand macro replacement when concatenating tokens or performing stringification 18PRE06-C. Enclose header files in an inclusion guard 21PRE07-C. Avoid using repeated question marks 22PRE08-C. Guarantee that header file names are unique 24PRE09-C. Do not replace secure functions with less secure functions 26PRE10-C. Wrap multistatement macros in a do-while loop 27PRE30-C. Do not create a universal character name through concatenation 29PRE31-C. Never invoke an unsafe macro with arguments containing assignment, increment, decrement, volatile access, or function call 30 Chapter 3: Declarations and Initialization (DCL) 33 DCL00-C. const-qualify immutable objects 35DCL01-C. Do not reuse variable names in subscopes 36DCL02-C. Use visually distinct identifiers 38DCL03-C. Use a static assertion to test the value of a constant expression 39DCL04-C. Do not declare more than one variable per declaration 42DCL05-C. Use type definitions to improve code readability 44DCL06-C. Use meaningful symbolic constants to represent literal values in program logic 45DCL07-C. Include the appropriate type information in function declarators 51DCL08-C. Properly encode relationships in constant definitions 54DCL09-C. Declare functions that return an errno error code with a return type of errno_t 57DCL10-C. Maintain the contract between the writer and caller of variadic functions 59DCL11-C. Understand the type issues associated with variadic functions 62DCL12-C. Implement abstract data types using opaque types 64DCL13-C. Declare function parameters that are pointers to values not changed by the function as const 66DCL14-C. Do not make assumptions about the order of global variable initialization across translation units 69DCL15-C. Declare objects that do not need external linkage with the storage-class specifier static 70DCL30-C. Declare objects with appropriate storage durations 72DCL31-C. Declare identifiers before using them 74DCL32-C. Guarantee that mutually visible identifiers are unique 78DCL33-C. Ensure that restrict-qualified source and destination pointers in function arguments do not reference overlapping objects 80DCL34-C. Use volatile for data that cannot be cached 82DCL35-C. Do not convert a function using a type that does not match the function definition 84DCL36-C. Do not declare an identifier with conflicting linkage classifications 87 Chapter 4: Expressions (EXP) 91 EXP00-C. Use parentheses for precedence of operation 93EXP01-C. Do not take the size of a pointer to determine the size of the pointed-to type 95EXP02-C. Be aware of the short-circuit behavior of the logical AND and OR operators 96EXP03-C. Do not assume the size of a structure is the sum of the sizes of its members 98EXP04-C. Do not perform byte-by-byte comparisons between structures 100EXP05-C. Do not cast away a const qualification 102EXP06-C. Operands to the sizeof operator should not contain side effects 104EXP07-C. Do not diminish the benefits of constants by assuming their values in expressions 105EXP08-C. Ensure pointer arithmetic is used correctly 107EXP09-C. Use sizeof to determine the size of a type or variable 109EXP10-C. Do not depend on the order of evaluation of subexpressions or the order in which side effects take place 111EXP11-C. Do not apply operators expecting one type to data of an incompatible type 114EXP12-C. Do not ignore values returned by functions 118EXP30-C. Do not depend on order of evaluation between sequence points 119EXP31-C. Avoid side effects in assertions 122EXP32-C. Do not cast away a volatile qualification 123EXP33-C. Do not reference uninitialized memory 124EXP34-C. Ensure a null pointer is not dereferenced 128EXP35-C. Do not access or modify the result of a function call after a subsequent sequence point 129EXP36-C. Do not convert pointers into more strictly aligned pointer types 131EXP37-C. Call functions with the arguments intended by the API 133EXP38-C. Do not call offsetof() on bit-field members or invalid types 135 Chapter 5: Integers (INT) 139 INT00-C. Understand the data model used by your implementation(s) 141INT01-C. Use rsize_t or size_t for all integer values representing the size of an object 145INT02-C. Understand integer conversion rules 149INT03-C. Use a secure integer library 153INT04-C. Enforce limits on integer values originating from untrusted sources 155INT05-C. Do not use input functions to convert character data if they cannot handle all possible inputs 157INT06-C. Use strtol() or a related function to convert a string token to an integer 159INT07-C. Use only explicitly signed or unsigned char type for numeric values 162INT08-C. Verify that all integer values are in range 164INT09-C. Ensure enumeration constants map to unique values 167INT10-C. Do not assume a positive remainder when using the % operator 168INT11-C. Take care when converting from pointer to integer or integer to pointer 170INT12-C. Do not make assumptions about the type of a plain int bit-field when used in an expression 172INT13-C. Use bitwise operators only on unsigned operands 174INT14-C. Avoid performing bitwise and arithmetic operations on the same data 175INT15-C. Use intmax_t or uintmax_t for formatted I/O on programmer-defined integer types 178INT30-C. Ensure that unsigned integer operations do not wrap 181INT31-C. Ensure that integer conversions do not result in lost or misinterpreted data 186INT32-C. Ensure that operations on signed integers do not result in overflow 191INT33-C. Ensure that division and modulo operations do not result in divide-by-zero errors 201INT34-C. Do not shift a negative number of bits or more bits than exist in the operand 203INT35-C. Evaluate integer expressions in a larger size before comparing or assigning to that size 207 Chapter 6: Floating Point (FLP) 211 FLP00-C. Understand the limitations of floating-point numbers 212FLP01-C. Take care in rearranging floating-point expressions 214FLP02-C. Consider avoiding floating-point numbers when precise computation is needed 215FLP03-C. Detect and handle floating-point errors 218FLP30-C. Do not use floating-point variables as loop counters 224FLP31-C. Do not call functions expecting real values with complex values 226FLP32-C. Prevent or detect domain and range errors in math functions 227FLP33-C. Convert integers to floating point for floating-point operations 234FLP34-C. Ensure that floating-point conversions are within range of the new type 236 Chapter 7: Arrays (ARR) 241 ARR00-C. Understand how arrays work 242ARR01-C. Do not apply the sizeof operator to a pointer when taking the size of an array 245ARR02-C. Explicitly specify array bounds, even if implicitly defined by an initializer 248ARR30-C. Guarantee that array indices are within the valid range 250ARR31-C. Use consistent array notation across all source files 251ARR32-C. Ensure size arguments for variable length arrays are in a valid range 254ARR33-C. Guarantee that copies are made into storage of sufficient size 255ARR34-C. Ensure that array types in expressions are compatible 258ARR35-C. Do not allow loops to iterate beyond the end of an array 259ARR36-C. Do not subtract or compare two pointers that do not refer to the same array 261ARR37-C. Do not add or subtract an integer to a pointer to a non-array object 263ARR38-C. Do not add or subtract an integer to a pointer if the resulting value does not refer to a valid array element 265 Chapter 8: Characters and Strings (STR) 271 STR00-C. Represent characters using an appropriate type 273STR01-C. Adopt and implement a consistent plan for managing strings 275STR02-C. Sanitize data passed to complex subsystems 276STR03-C. Do not inadvertently truncate a null-terminated byte string 280STR04-C. Use plain <cod>char</code> for characters in the basic character set 282STR05-C. Use pointers to const when referring to string literals 284STR06-C. Do not assume that strtok() leaves the parse string unchanged 286STR07-C. Use TR 24731 for remediation of existing string manipulation code 288STR08-C. Use managed strings for development of new string manipulation code 291STR30-C. Do not attempt to modify string literals 293STR31-C. Guarantee that storage for strings has sufficient space for character data and the null terminator 294STR32-C. Null-terminate byte strings as required 299STR33-C. Size wide character strings correctly 303STR34-C. Cast characters to unsigned types before converting to larger integer sizes 305STR35-C. Do not copy data from an unbounded source to a fixed-length array 307STR36-C. Do not specify the bound of a character array initialized with a string literal 312STR37-C. Arguments to character-handling functions must be representable as an unsigned char 314 Chapter 9: Memory Management (MEM) 317 MEM00-C. Allocate and free memory in the same module at the same level of abstraction 319MEM01-C. Store a new value in pointers immediately after free() 322MEM02-C. Immediately cast the result of a memory allocation function call into a pointer to the allocated type 324MEM03-C. Clear sensitive information stored in reusable resources returned for reuse 328MEM04-C. Do not perform zero-length allocations 332MEM05-C. Avoid large stack allocations 335MEM06-C. Ensure that sensitive data is not written out to disk 338MEM07-C. Ensure that the arguments to calloc(), when multiplied, can be represented as a size_t 342MEM08-C. Use realloc() only to resize dynamically allocated arrays 343MEM09-C. Do not assume memory allocation routines initialize memory 346MEM10-C. Use a pointer validation function 348MEM30-C. Do not access freed memory 351MEM31-C. Free dynamically allocated memory exactly once 353MEM32-C. Detect and handle memory allocation errors 355MEM33-C. Use the correct syntax for flexible array members 358MEM34-C. Only free memory allocated dynamically 360MEM35-C. Allocate sufficient memory for an object 362 Chapter 10: Input/Output (FIO) 367 FIO00-C. Take care when creating format strings 370FIO01-C. Be careful using functions that use file names for identification 372FIO02-C. Canonicalize path names originating from untrusted sources 374FIO03-C. Do not make assumptions about fopen() and file creation 383FIO04-C. Detect and handle input and output errors 386FIO05-C. Identify files using multiple file attributes 389FIO06-C. Create files with appropriate access permissions 394FIO07-C. Prefer fseek() to rewind() 398FIO08-C. Take care when calling remove() on an open file 399FIO09-C. Be careful with binary data when transferring data across systems 401FIO10-C. Take care when using the rename() function 403FIO11-C. Take care when specifying the mode parameter of fopen() 407FIO12-C. Prefer setvbuf() to setbuf() 408FIO13-C. Never push back anything other than one read character 409FIO14-C. Understand the difference between text mode and binary mode with file streams 411FIO15-C. Ensure that file operations are performed in a secure directory 413FIO16-C. Limit access to files by creating a jail 418FIO30-C. Exclude user input from format strings 421FIO31-C. Do not simultaneously open the same file multiple times 424FIO32-C. Do not perform operations on devices that are only appropriate for files 426FIO33-C. Detect and handle input output errors resulting in undefined behavior 431FIO34-C. Use int to capture the return value of character I/O functions 436FIO35-C. Use feof() and ferror() to detect end-of-file and file errors when sizeof(int) == sizeof(char) 438FIO36-C. Do not assume a new-line character is read when using fgets() 440FIO37-C. Do not assume character data has been read 442FIO38-C. Do not use a copy of a FILE object for input and output 443FIO39-C. Do not alternately input and output from a stream without an intervening flush or positioning call 444FIO40-C. Reset strings on fgets() failure 446FIO41-C. Do not call getc() or putc() with stream arguments that have side effects 448FIO42-C. Ensure files are properly closed when they are no longer needed 450FIO43-C. Do not create temporary files in shared directories 454FIO44-C. Only use values for fsetpos() that are returned from fgetpos() 464 Chapter 11: Environment (ENV) 467 ENV00-C. Do not store the pointer to the string returned by getenv() 468ENV01-C. Do not make assumptions about the size of an environment variable 474ENV02-C. Beware of multiple environment variables with the same effective name 475ENV03-C. Sanitize the environment when invoking external programs 478ENV04-C. Do not call system() if you do not need a command processor 482ENV30-C. Do not modify the string returned by getenv() 487ENV31-C. Do not rely on an environment pointer following an operation that may invalidate it 489ENV32-C. No atexit handler should terminate in any way other than by returning 494 Chapter 12: Signals (SIG) 499 SIG00-C. Mask signals handled by noninterruptible signal handlers 500SIG01-C. Understand implementation-specific details regarding signal handler persistence 503SIG02-C. Avoid using signals to implement normal functionality 507SIG30-C. Call only asynchronous-safe functions within signal handlers 511SIG31-C. Do not access or modify shared objects in signal handlers 517SIG32-C. Do not call longjmp() from inside a signal handler 519SIG33-C. Do not recursively invoke the raise() function 523SIG34-C. Do not call signal() from within interruptible signal handlers 526 Chapter 13: Error Handling (ERR) 531 ERR00-C. Adopt and implement a consistent and comprehensive error-handling policy 533ERR01-C. Use ferror() rather than errno to check for FILE stream errors 535ERR02-C. Avoid in-band error indicators 537ERR03-C. Use runtime-constraint handlers when calling functions defined by TR 24731-1 541ERR04-C. Choose an appropriate termination strategy 544ERR05-C. Application-independent code should provide error detection without dictating error handling 549ERR06-C. Understand the termination behavior of assert() and abort() 556ERR30-C. Set errno to zero before calling a library function known to set errno, and check errno only after the function returns a value indicating failure 558ERR31-C. Do not redefine errno 563ERR32-C. Do not rely on indeterminate values of errno 564 Chapter 14: Miscellaneous (MSC) 569 MSC00-C. Compile cleanly at high warning levels 570MSC01-C. Strive for logical completeness 572MSC02-C. Avoid errors of omission 574MSC03-C. Avoid errors of addition 576MSC04-C. Use comments consistently and in a readable fashion 578MSC05-C. Do not manipulate time_t typed values directly 580MSC06-C. Be aware of compiler optimization when dealing with sensitive data 582MSC07-C. Detect and remove dead code 585MSC08-C. Library functions should validate their parameters 588MSC09-C. Character encoding: use subset of ASCII for safety 590MSC10-C. Character encoding: UTF-8-related issues 594MSC11-C. Incorporate diagnostic tests using assertions 597MSC12-C. Detect and remove code that has no effect 598MSC13-C. Detect and remove unused values 600MSC14-C. Do not introduce unnecessary platform dependencies 602MSC15-C. Do not depend on undefined behavior 604MSC30-C. Do not use the rand() function for generating pseudorandom numbers 607MSC31-C. Ensure that return values are compared against the proper type 610 Appendix: POSIX (POS) 613 POS00-C. Avoid race conditions with multiple threads 615POS01-C. Check for the existence of links 617POS02-C. Follow the principle of least privilege 620POS30-C. Use the readlink() function properly 623POS31-C. Do not unlock or destroy another thread’s mutex 625POS32-C. Include a mutex when using bit-fields in a multithreaded environment 626POS33-C. Do not use vfork() 629POS34-C. Do not call putenv() with a pointer to an automatic variable as the argument 631POS35-C. Avoid race conditions while checking for the existence of a symbolic link 633POS36-C. Observe correct revocation order while relinquishing privileges 636POS37-C. Ensure that privilege relinquishment is successful 637 Glossary 643References 647Index 659
An essential element of secure coding in the C programming language is well-documented and enforceable coding standards. Coding standards encourage programmers to follow a uniform set of guidelines determined by the requirements of the project and organization, rather than by the programmer's familiarity or preference. Once established, these standards can be used as a metric to evaluate source code (using manual or automated processes).
The CERT C Secure Coding Standard provides guidelines for secure coding in the C programming language. The goal of these guidelines is to eliminate insecure coding practices and undefined behaviors that can lead to exploitable vulnerabilities. The application of the secure coding standard will lead to higher-quality systems that are robust and more resistant to attack.
The CERT C Secure Coding Standard was developed over a period of two and a half years as a community effort and involved the efforts of 226 contributors and reviewers including a half-dozen active members of the ISO/IEC WG14 international standardization working group for the programming language C, the Chairman and Vice Chairman of PL22.11 (ANSI "C"), representatives from the Open Group, USENIX, Microsoft, and numerous other companies and organizations. Drafts of The CERT C Secure Coding Standard were twice reviewed by ISO/IEC WG14 and subjected to the scrutiny of the public including members of the Association of C and C++ Users (ACCU) and the comp.lang.c news group.
The results of this effort are 89 rules and 132 recommendations for secure coding in the C programming language. Most of these guidelines come complete with insecure (non-compliant) code examples,and secure (compliant solutions). The CERT C Secure Coding Standards are supported by training available from the Software Engineering Institute and other licensed partners. A number of source code analysis tools are available to automatically detect violations of CERT Secure Coding Standard rules and recommendations, including Compass/ROSE which is freely available from Lawrence Livermore National Laboratory and CERT.The Demand for Secure Software
The Morris worm incident, which brought ten percent of Internet systems to a halt in November 1988, resulted in a new and acute awareness of the need for secure software systems. Twenty years later, many security analysts, software developers, software users, and policy makers are asking the question "Why isn't software more secure?"
The first problem is that the term software security, as it is used today, is meaningless. I have attempted to define this term, as have others, but there is no generally accepted definition. Why does this matter?
There are a variety of reasons given for why software is not more secure, such as the tools are inadequate, programmers lack sufficient training, and schedules are too short. But these are all solvable problems. The root cause of the issue lies elsewhere.
The reason more software is not more secure is because there is no demand for secure software. In simple terms, if one vendor offers a product that has more features, better performance, and is available today and another vendor offers a secure product that has less features, not quite as good performance, and will be available in six months, there is really no question as to which product customers will buy, and vendors know this.
So why don't customers buy secure products? Again, it is because the word "secure" is meaningless in this context. Why would a customer pass up tangible benefits to buy a product that has an ill-defined and intangible property?
This is the problem addressed by the CERT C Secure Coding Standard. This book contains 89 rules and 132 recommendations for producing secure code. While the application of these rules and recommendations does not guarantee the security of a software system, it does tell you a great deal about the quality and security of the code. It tells you that the software was developed to a set of industry standard rules and recommendations that were developed by the leading experts in the field. It tells you that a tremendous amount of time and effort went into producing code that is free from the common coding errors that have resulted in numerous vulnerabilities that have been reported to and published by the CERT Coordination Center over the past two decades. It tells you that the software developers who produced the code have done so with a real knowledge of the types of vulnerabilities that can exist and the exploits that can be used against them, and consequently have developed the software with a real security mindset.
So, the small problem we have set out to address in this book is to change the market dynamic for developing and purchasing software systems. By producing an actionable definition of software security for C language programs--compliance with the rules and recommendations in this standard--we have defined a mechanism by which customers can demand secure software systems and vendors can comply. Furthermore, the concept of a secure system now has value because the word "secure" has meaning.History
I have participated in C language standardization efforts for the past several years as the Carnegie Mellon University representative to INCITS J11 (now PL22.11) and as a technical expert at ISO/IEC WG14 (the international standardization working group for the programming language C). The first WG14 meeting I attended was held in April 2005 in Lillehammer, Norway, where we discussed a proposal for a Specification for Secure C Library Functions. That specification would eventually be published as TR 24731-1, Extensions to the C Library - Part 1: Bounds Checking Interfaces .
Although the TR 24731-1 are "more secure," they are still susceptible to reuse. Consequently, a separate effort was started at the WG14 meeting to develop PDTR 24731-2, Extensions to the C Library - Part II: Dynamic Allocation Functions ISO/IEC PDTR 24731-2, consisting primarily of existing POSIX and Linux functions. At that time, I thought it would make sense to develop a managed string library to provide a set of dynamic allocation functions with a consistent API and propose it to WG14 for standardization. One year later in Berlin, Germany, I was told that I had come up with a good technical solution but there was no customer demand for such a library.
Minutes later, during a break, I had a "Mean Joe Green" moment in the hallway when Tom Plum approach me and suggested that perhaps the C programming community would benefit from CERT developing a secure coding standard. I immediately saw the wisdom of this proposal. The C99 standard is a authoritative document, but the audience for it is primarily compiler implementers and, as been noted by many, its language is obscure and often impenetrable. A secure coding standard would be primarily targeted towards C language programmers and would provide actionable guidance on how to code securely in the language.Community Development Process
The development of a secure coding standard for any programming language is a difficult undertaking that requires significant community involvement. The following development process has been used to create this standard:
- Rules and recommendations for a coding standard are solicited from the communities involved in the development and application of each programming language, including the formal or de facto standard bodies responsible for the documented standard and user groups.
- These rules and recommendations are edited by members of the CERT technical staff and industry experts for content and style on the CERT Secure Coding Standards wiki at www.securecoding.cert.org.
- The user community reviews and comments on the publicly posted content using threaded discussions and other communication tools. If a consensus develops that the rule or recommendation is appropriate and correct, it is incorporated into an officially released version of the secure coding standard. If the rule does not achieve consensus, it is moved to a special section called "The Void". From here, it may be resurrected (usually in a altered form) or removed.
This development approach has been highly successful with numerous individuals and organizations contributing their time and expertise to the project. As a result of this process, The CERT C Secure Coding Standard has achieved a level of completeness and thoroughness that would not otherwise be achievable by a single author, or even a small team of authors.
The main disadvantage of developing a secure coding standard on a wiki is that the content is constantly evolving. This is great if you want the latest information, and you ware willing to entertain the possibility that a recent change has not yet been fully vetted. However, many software development organizations require a final document before they can commit to complying with a (fixed) set of rules and recommendations. This book serves that purpose, as Version 1.0 of the CERT C Secure Coding Standard.
With the production of the manuscript for this book in June of 2008, Version 1.0 (this book) and the wiki versions of the Secure Coding Standard began to diverge. Because both the C programming language and our knowledge of how to use it securely is still evolving, CERT will continue to evolve the CERT C Secure Coding Standard" on the secure coding wiki. These changes may then be incorporated into future, officially released versions of this standard.Purpose
The CERT C Secure Coding Standard provides developers with guidelines for secure coding in the C programming language. These guidelines serve a variety of purposes. First, they enumerate common errors in C language programming that can lead to software defects, security flaws, and software vulnerabilities. These are all errors for which a conforming compiler is not required by the standard to issue a fatal diagnostic. In other words, the compiler will generate an executable, frequently without warning, and the resulting code executable will contain flaws that may make it vulnerable to attack.
Second, this coding standard provides recommendations for how to produce secure code. Failure to comply with these recommendations does not necessarily mean that the software is insecure, but if followed these recommendations can be powerful tools in eliminating vulnerabilities from software.
Third, this coding standard identifies non-portable coding practices. Portability is not a strict requirement of security, but non-portable assumptions in code often result in vulnerabilities when code is ported to platforms for which these assumptions are no longer valid.
Guidelines are classified as either rules or recommendations. Guidelines are defined to be rules when all of the following conditions are met:
- Violation of the coding practice is likely to result in a security flaw that may result in an exploitable vulnerability.
- There is a denumerable set of conditions for which violating the coding practice is necessary to ensure correct behavior.
- Conformance to the coding practice can be determined through automated analysis, formal methods, or manual inspection techniques.
Implementation of the secure coding rules defined in this standard are necessary (but not sufficient) to ensure the security of software systems developed in the C programming language.
Recommendations are guidelines or suggestions. Guidelines are defined to be recommendations when all of the following conditions are met:
- Application of the coding practice is likely to improve system security.
- One or more of the requirements necessary for a coding practice to be considered a rule cannot be met.
The set of recommendations that a particular development effort adopts depends on the security requirements of the final software product. Projects with high-security requirements can dedicate more resources to security and are consequently likely to adopt a larger set of recommendations.
To ensure that the source code conforms to this secure coding standard, it is necessary to have measures in place that check for rules violations. The most effective means of achieving this is to use one or more static analysis tools. Where a rule cannot be checked by a tool, then a manual review is required.Scope
The CERT C Programming Language Secure Coding Standard was developed specifically for versions of the C programming language defined by
• ISO/IEC 9899:1999 Programming Languages -- C, Second Edition
• Technical corrigenda TC1, TC2, and TC3
• ISO/IEC TR 24731-1 Extensions to the C Library, Part I: Bounds-checking interfaces
• ISO/IEC WDTR 24731-2 Extensions to the C Library, Part II: Dynamic Allocation Functions
Most of the material included in this standard can also be applied to earlier versions of the C programming language.
Rules and recommendations included in this CERT C Programming Language Secure Coding Standard are designed to be operating system and platform independent. However, the best solutions to secure coding problems are often platform specific. In most cases, this standard provides appropriate compliant solutions for POSIX-compliant and Windows operating systems. In many cases, compliant solutions have also been provided for specific platforms such as Linux or OpenBSD. Occasionally, we also point out implementation-specific behaviors when these behaviors are of interest.Rationale
A secure coding standard for the C programming language can create the highest value for the longest period of time by focusing on C99 and the relevant post-C99 technical reports. In addition, because considerably more money and effort is devoted to developing new code than maintaining existing code, the highest return on investment comes from influencing programmers who are developing new code. Maintaining existing code is still an important concern, however.
The C standard documents existing practice where possible. That is, most features must be tested in an implementation before being included in the standard. The CERT C secure coding standard has a different purpose. When existing practice serves this purpose, that is fine, but the goal is to create a new set of best practices, and that includes introducing some concepts that are not yet widely known. To put it a different way, the CERT C secure coding guidelines are attempting to drive change rather than just document it.
For example, the C library technical report, part 1 (TR 24731-1) is gaining support, but at present is only implemented by a few vendors. It introduces functions such as memcpy_s(), which serve the purpose of security by adding the destination buffer size to the API. A forward-looking document could not reasonably ignore these simply because they are not yet widely implemented.
C99 is more widely implemented, but even if it were not yet, it is the direction in which the industry is moving. Developers of new C code, especially, need guidance that is usable on and makes the best use of the compilers and tools that are now being developed and are being supported into the future.
Some vendors have extensions to C, and some also have implemented only part of the C standard before stopping development. Consequently, it is not possible to back up and only discuss C95, or C90. The vendor support equation is too complicated to draw a line and say that a certain compiler supports exactly a certain standard. Whatever demarcation point is selected, different vendors are on opposite sides of it for different parts of the language. Supporting all possibilities would require testing the cross product of each compiler with each language feature. Consequently, we have selected a demarcation point that is the most recent in time, so that the rules and recommendations defined by the standard will be applicable for as long as possible. As a result of the variations in support, source-code portability is enhanced when the programmer uses only the features specified by C90. This is one of many trade-offs between security and portability inherent to C language programming.
The value of forward looking information increases with time before it starts to decrease. The value of backward-looking information starts to decrease immediately.
For all these reasons, the priority of this standard is to support new code development using C99 and the post-C99 technical reports. A close-second priority is supporting remediation of old code using C99 and the technical reports.
This standard does try to make contributions to support older compilers when these contributions can be significant and doing so does not compromise other priorities. The intent is not to capture all deviations from the standard but only a few important ones.Issues Not Addressed
There are a number of issues not addressed by this secure coding standard.
• Coding Style. Coding style issues are subjective, and it has proven impossible to develop a consensus on appropriate style guidelines. Consequently, the CERT C Secure Coding standard does not require any particular coding style to be enforced but only that the user defines style guidelines and apply these guidelines consistently. The easiest way to consistently apply a coding style is with the use of a code formatting tool. Many interactive development environments (IDEs) provide such capabilities.
• Tools. As a federally funded research and development center (FFRDC), the SEI is not in a position to recommend particular vendors or tools to enforce the restrictions adopted. The user of this document is free to choose tools, and vendors are encouraged to provide tools to enforce the rules.
• Controversial Rules. In general, the CERT secure coding standards try to avoid the inclusion of controversial rules that lack a broad consensus.Who Should Read This Book
The CERT C Secure Coding Standard is primarily intended for developers of C language programs. While security is an important for Internet-facing systems, for example, it is also an important concern for any software component that may be included or deployed as part of a secure software system. With systems increasingly being composed of software components, or even other systems, it is difficult to identify situations in which software is guaranteed not to be used in another context, which perhaps has more stringent security requirements.
This book is also useful for C language programmers who don't realize they are interested in security as most of these guidelines have practical applications for achieving other quality attributes such as safety, reliability, dependability, robustness, availability, maintainability.
While not intended for C++ programmers, this book may also be of some value because the vast majority of issues identified for C language programs are also issues in C++ programmers, although in many cases the solutions are different.
Another group of individuals that can benefit from reading this book are the members of the ISO/IEC WG14 (the international standardization working group for the programming language C) as they consider software security requirements for the new major revision of the C language standard (C1X) currently being developed.How This Book is Organized
This book is organized into an introductory chapter, thirteen chapters each containing rules and recommendations in a particular topic area, and an appendix containing rules and recommendations for POSIX to demonstrate how this secure coding standard can be customized for particular environments.
Chapter 1, Introduction
Chapter 2, Preprocessor (PRE)
Chapter 3, Declarations and Initialization (DCL)
Chapter 4, Expressions (EXP)
Chapter 5, Integers (INT)
Chapter 6, Floating Point (FLP)
Chapter 7, Arrays (ARR)
Chapter 8, Characters and Strings (STR)
Chapter 9, Memory Management (MEM)
Chapter 10, Input Output (FIO)
Chapter 11, Environment (ENV)
Chapter 12, Signals (SIG)
Chapter 13, Error Handling (ERR)
Chapter 14, Miscellaneous (MSC)
Appendix A, POSIX (POS)
The POSIX appendix is non-normative and not a prescriptive part of the standard.Notes to the Reader
As noted, the CERT C Secure coding standard is organized into chapters, each containing a set of guidelines in a particular topic area. Each guideline in this standard has a unique identifier, which is included in the title.
Most guidelines have a consistent structure. The title of the guidelines and the introductory paragraphs define the rule or recommendation. This is typically followed by one or more pairs of non-compliant code examples and compliant solutions. Each guideline also includes a risk assessment and a list of appropriate references (where applicable). Guidelines will also include a table of related vulnerabilities, where identified.Identifiers
These identifiers consist of three parts:
• A three-letter mnemonic representing the section of the standard
• A two-digit numeric value in the range of 00-99
• The letter "A" or "C" to indicate whether the coding practice is an advisory recommendation or a compulsory rule
The three-letter mnemonic can be used to group similar guidelines and to indicate to which category a guideline belongs.
The numeric value is used to give each guideline a unique identifier. Numeric values in the range of 00-29 are reserved for recommendations, while values in the range of 30-99 are reserved for rules.
The letter "A" or "C" in the identifier is not required to uniquely identify guideline. It is used only to provide a clear indication of whether the coding practice is an advisory recommendation or a compulsory rule.Non-Compliant Code Exemples and Compliant Solutions
Non-compliant code examples are examples of insecure code that violate the guideline under discussion. It is important to note that these are only examples, and eliminating all occurrences of the example does not necessarily mean that your code is now compliant with the guideline.
The non-compliant code examples are typically followed by compliant solutions, which are examples of how the logic from the corresponding non-compliant code example can be coded in a secure, compliant manner. Except where noted, non-compliant code examples should only contain a violation of the rule under discussion. Compliant solutions should comply with all secure coding rules, but may on occasion fail to comply with a recommendation as noted.Risk Assessment
Each guideline contains a risk assessment section, which attempts to quantify and qualify the risk of violating each guideline. This information is intended primarily for remediation projects to help prioritize repairs, as it is assumed that new development efforts will conform with the entire standard.
Each rule and recommendation has an assigned priority. Priorities are assigned using a metric based on Failure Mode, Effects, and Criticality Analysis (FMECA). Three values are assigned for each rule on a scale of 1 to 3 for
• Severity - how serious are the consequences of the rule being ignored
1 = low (denial-of-service attack, abnormal termination)
2 = medium (data integrity violation, unintentional information disclosure)
3 = high (run arbitrary code)
• Likelihood - how likely is it that a flaw introduced by ignoring the rule could lead to an exploitable vulnerability
1 = unlikely
2 = probable
3 = likely
• Remediation cost - how expensive is it to comply with the rule
1 = high (manual detection and correction)
2 = medium (automatic detection and manual correction)
3 = low (automatic detection and correction)
The three values are then multiplied together for each rule. This product provides a measure that can be used in prioritizing the application of the rules. These products range from 1 to 27. Rules and recommendations with a priority in the range of 1-4 are level 3 rules, 6-9 are level 2, and 12-27 are level 1. As a result, it is possible to claim level 1, level 2, or complete compliance (level 3) with a standard by implementing all rules in a level.
Recommendations are not compulsory and are provided for information purposes only.References
Guidelines include frequent references to the vulnerability notes in CERT's Coordination Center Vulnerability Notes Database, CWE IDs in MITRE's Common Weakness Enumeration (CWE) MITRE 07, and CVE numbers from MITRE's Common Vulnerabilities and Exposures (CVE).
You can create a unique URL to get more information on any of these topics by appending the relevant ID to the end of a fixed string. For example, to find more information about:
• VU#551436, "Mozilla Firefox SVG viewer vulnerable to integer overflow," you can append 551436 to https://www.kb.cert.org/vulnotes/id/ and enter the resulting URL in your browser: https://www.kb.cert.org/vulnotes/id/551436
• CWE ID 192, "Integer Coercion Error" you can append "192.html" to "http://cwe.mitre.org/data/definitions/" and enter the resulting URL in your browser: http://cwe.mitre.org/data/definitions/192.html
• CVE-2006-1174, you can append "CVE-2006-1174" to "http://cve.mitre.org/cgi-bin/cvename.cgi?name=" and enter the resulting URL in your browser: http://cve.mitre.org/cgi-bin/cvename.cgi?name= CVE-2006-1174
Guidelines are frequently correlated with language vulnerabilities in ISO/IEC PDTR 24772. Information Technology -- Programming Languages -- Guidance to Avoiding Vulnerabilities in Programming Languages through Language Selection and UseRelated Vulnerabilities
Wherever possible, we have tried to link the rules and recommendations in this secure coding standard to violations of actual vulnerabilities published in the CERT Coordination Center Vulnerability Notes Database. New links are continually added. To find the latest list of related vulnerabilities, enter the following URL:
where "XXXNN-X" is the ID of the rule or recommendation for which you are searching.
These tables consist of four fields: metric, ID, date public, and name.Vulnerability Metric
The CERT vulnerability metric value is a number between 0 and 180 that assigns an approximate severity to the vulnerability. This number considers several factors:
• Is information about the vulnerability widely available or known?
• Is the vulnerability being exploited in incidents reported to CERT or other incident response teams?
• Is the Internet infrastructure (e.g., routers, name servers, critical Internet protocols) at risk because of this vulnerability?
• How many systems on the Internet are at risk from this vulnerability?
• What is the impact of exploiting the vulnerability?
• How easy is it to exploit the vulnerability?
• What are the preconditions required to exploit the vulnerability?
Because the questions are answered with approximate values based on our own judgments and may differ significantly from one site to another, you should not rely too heavily on the metric for prioritizing response to vulnerabilities. Rather, this metric may be useful for separating the serious vulnerabilities from the larger number of less severe vulnerabilities described in the database. Because the questions are not all weighted equally, the resulting score is not linear (that is, a vulnerability with a metric of 40 is not twice as severe as one with a metric of 20).
An alternative vulnerability severity metric is the Common Vulnerability Scoring System (CVSS).Vulnerability ID
Vulnerability ID numbers are assigned at random to uniquely identify a vulnerability. These IDs are four to six digits long, and are usually prefixed with "VU#" to mark them as vulnerability IDs.Date Public
This is the date on which the vulnerability was first known to the public, to the best of our knowledge. Usually this date is when the Vulnerability Note was first published, when an exploit was first discovered, when the vendor first distributed a patch publicly, or when a description of the vulnerability was posted to a public mailing list. By default, this date is set to be our Vulnerability Note publication date.Vulnerability Name
The vulnerability name is a short description that summarizes the nature of the problem and the affected software product. While the name may include a clause describing the impact of the vulnerability, most names are focused on the nature of the defect that caused the problem to occur.