Pseudorandom number generators use mathematical algorithms to produce a sequence of numbers with good statistical properties, but the numbers produced are not genuinely random.
The C Standard rand()
function makes no guarantees as to the quality of the random sequence produced. The numbers generated by some implementations of rand()
have a comparatively short cycle and the numbers can be predictable. Applications that have strong pseudorandom number requirements must use a generator that is known to be sufficient for their needs.
Noncompliant Code Example
The following noncompliant code generates an ID with a numeric part produced by calling the rand()
function. The IDs produced are predictable and have limited randomness.
#include <stdio.h> #include <stdlib.h> enum { len = 12 }; void func(void) { /* * id will hold the ID, starting with the characters * "ID" followed by a random integer. */ char id[len]; int r; int num; /* ... */ r = rand(); /* Generate a random integer */ num = snprintf(id, len, "ID%-d", r); /* Generate the ID */ /* ... */ }
Compliant Solution (POSIX)
This compliant solution replaces the rand()
function with the POSIX random()
function:
#include <stdio.h> #include <stdlib.h> #include <time.h> enum { len = 12 }; void func(void) { /* * id will hold the ID, starting with the characters * "ID" followed by a random integer. */ char id[len]; int r; int num; /* ... */ struct timespec ts; if (timespec_get(&ts, TIME_UTC) == 0) { /* Handle error */ } srandom(ts.tv_nsec ^ ts.tv_sec); /* Seed the PRNG */ /* ... */ r = random(); /* Generate a random integer */ num = snprintf(id, len, "ID%-d", r); /* Generate the ID */ /* ... */ }
The POSIX random()
function is a better pseudorandom number generator. Although on some platforms the low dozen bits generated by rand()
go through a cyclic pattern, all the bits generated by random()
are usable. The rand48
family of functions provides another alternative for pseudorandom numbers.
Although not specified by POSIX, arc4random()
is another possibility for systems that support it. The arc4random(3)
manual page [OpenBSD] states
... provides higher quality of data than those described in rand(3), random(3), and drand48(3).
To achieve the best random numbers possible, an implementation-specific function must be used. When unpredictability is crucial and speed is not an issue, as in the creation of strong cryptographic keys, use a true entropy source, such as /dev/random
, or a hardware device capable of generating random numbers. The /dev/random
device can block for a long time if there are not enough events going on to generate sufficient entropy.
Compliant Solution (Windows)
On Windows platforms, the BCryptGenRandom()
function can be used to generate cryptographically strong random numbers. The Microsoft Developer Network BCryptGenRandom()
reference [MSDN] states:
The default random number provider implements an algorithm for generating random numbers that complies with the NIST SP800-90 standard, specifically the CTR_DRBG portion of that standard.
#include <Windows.h> #include <bcrypt.h> #include <stdio.h> #pragma comment(lib, "Bcrypt") void func(void) { BCRYPT_ALG_HANDLE Prov; int Buffer; if (!BCRYPT_SUCCESS( BCryptOpenAlgorithmProvider(&Prov, BCRYPT_RNG_ALGORITHM, NULL, 0))) { /* handle error */ } if (!BCRYPT_SUCCESS(BCryptGenRandom(Prov, (PUCHAR) (&Buffer), sizeof(Buffer), 0))) { /* handle error */ } printf("Random number: %d\n", Buffer); BCryptCloseAlgorithmProvider(Prov, 0); }
Risk Assessment
The use of the rand()
function can result in predictable random numbers.
Rule | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
MSC30-C | Medium | Unlikely | Low | P6 | L2 |
Automated Detection
Tool | Version | Checker | Description |
---|---|---|---|
Astrée | 24.04 | stdlib-use-rand | Fully checked |
Axivion Bauhaus Suite | 7.2.0 | CertC-MSC30 | |
Clang | 4.0 (prerelease) | cert-msc30-c | Checked by clang-tidy |
CodeSonar | 8.1p0 | BADFUNC.RANDOM.RAND | Use of rand |
Compass/ROSE | |||
Coverity | 2017.07 | DONTCALL | Implemented - weak support |
Cppcheck Premium | 24.9.0 | premium-cert-msc30-c | Fully implemented |
1.2 | CC2.MSC30 | Fully implemented | |
Helix QAC | 2024.3 | C5022 C++5029 | |
Klocwork | 2024.3 | CERT.MSC.STD_RAND_CALL | |
LDRA tool suite | 9.7.1 | 44 S | Enhanced enforcement |
Parasoft C/C++test | 2023.1 | CERT_C-MSC30-a | Do not use the rand() function for generating pseudorandom numbers |
PC-lint Plus | 1.4 | 586 | Fully supported |
Polyspace Bug Finder | R2024a | CERT C: Rule MSC30-C | Checks for vulnerable pseudo-random number generator (rule fully covered) |
RuleChecker | 24.04 | stdlib-use-rand | Fully checked |
Related Vulnerabilities
Search for vulnerabilities resulting from the violation of this rule on the CERT website.
Related Guidelines
Key here (explains table format and definitions)
Taxonomy | Taxonomy item | Relationship |
---|---|---|
CERT C | MSC50-CPP. Do not use std::rand() for generating pseudorandom numbers | Prior to 2018-01-12: CERT: Unspecified Relationship |
CERT Oracle Secure Coding Standard for Java | MSC02-J. Generate strong random numbers | Prior to 2018-01-12: CERT: Unspecified Relationship |
CWE 2.11 | CWE-327, Use of a Broken or Risky Cryptographic Algorithm | 2017-05-16: CERT: Rule subset of CWE |
CWE 2.11 | CWE-330, Use of Insufficiently Random Values | 2017-06-28: CERT: Rule subset of CWE |
CWE 2.11 | CWE-338, Use of Cryptographically Weak Pseudo-Random Number Generator (PRNG) | 2017-06-28: CERT: Rule subset of CWE |
CWE 2.11 | CWE-676 | 2017-05-18: CERT: Rule subset of CWE |
CERT-CWE Mapping Notes
Key here for mapping notes
CWE-327 and MSC30-C
- CWE-327 forbids “broken or risky cryptographic algorithms” but does not specify what constitutes such an algo.
- Per CERT judgement, rand() qualifies, so:
- CWE-327 = Union( MSC30-C, list) where list =
- Invocation of broken/risky crypto algorithms besides rand()
CWE-338 and MSC30-C
CWE-338 = Union( MSC30-C, list) where list =
- Use of a weak PRNG besides standard C rand().
CWE-330 and MSC30-C
Independent( MSC30-C, MSC32-C, CON33-C)
CWE-330 = Union( MSC30-C, MSC32-C, CON33-C, list) where list = other improper use or creation of random values. (EG the would qualify)
MSC30-C, MSC32-C and CON33-C are independent, they have no intersections. They each specify distinct errors regarding PRNGs.
CWE-676 and MSC30-C
- Independent( ENV33-C, CON33-C, STR31-C, EXP33-C, MSC30-C, ERR34-C)
- MSC30-C implies that rand() is dangerous.
- CWE-676 = Union( MSC30-C, list) where list =
- Invocation of other dangerous functions, besides rand().
12 Comments
Jonathan Leffler
The compliant solution purportedly using that rand48 family of functions doesn't use any of them.
Arbob Ahmad
I think that the POSIX compliant solution would be better if it checked whether time(0) returned (time_t)-1 or not. This would better adhere to the earlier recommendation to check return values for errors. Otherwise, if an error occurs, an attacker could seed the random() function with -1 to determine the random numbers generated.
David Svoboda
Changed code to check for this. Also s/0/NULL/;
Not sure the check is necessary...my linux manpage sez time(2) only returns an error if the argument pointer is invalid. But just to be safe...
David Svoboda
Wietse Venema sez:
The (POSIX) compliant solution uses srandom(time(0)) for initialization,
but fails to mention that this makes the result of random() predictable.
The text should better distinguish between "random" (i.e. unpredictable)
and "pseudorandom". In the best case, current computers can produce
"pseudorandom" numbers but not "random" numbers".
Douglas A. Gwyn
Under compliant solution (POSIX), "While the low dozen bits generated by
rand()
go through a cyclic pattern, all the bits generated byrandom()
are usable." should be removed. rand() can be a high-quality generator, but that's not mandated. (In fact, POSIX mandated the medium-quality example shown in the C standard; instead of having a separate random() they could have required rand() to be better.)There should be advice to the effect: "High-quality design of pseudo-random number generators or cryptographic primitives requires an unusual degree of expertise, so you should generally adopt a proven implementation rather than trying to craft your own."
Alex Volkovitsky
Why is the remediation cost here "high"? s/rand/random/ is a quick automatic fix... this should be a cost of "low"
Justin Pincar
Maybe medium if you want to change it to the Windows one...? I agree it shouldn't be high.
David Svoboda
remediation cost is now 'low'.
versat
Microsoft marked CryptGenRandom as deprecated (replaced by BCryptGenRandom):
https://docs.microsoft.com/en-us/windows/desktop/api/wincrypt/nf-wincrypt-cryptgenrandom
And something else that i noticed:
Coverity marks random() as risky (but still recommends using the now deprecated CryptGenRandom()):
"Calling risky function (DC.WEAK_CRYPTO)dont_call: random should not be used for security related applications, as linear congruential algorithms are too easy to break.
Use a compliant random number generator, such as /dev/random or /dev/urandom on Unix-like systems, and CryptGenRandom on Windows."
Not really wrong or so but maybe interesting.
David Svoboda
Are you sure BCryptGenRandom is current? From MSDN's page, this function is only supported in Vista and Server 2008. Which aren't exactly modern OS's.
https://docs.microsoft.com/en-us/windows/desktop/api/Bcrypt/nf-bcrypt-bcryptgenrandom
Aaron Ballman
MSDN is documenting when those APIs were introduced; BCryptGenRandom() is current.
David Svoboda
With Aaron Ballman's help, I have replaced the CryptGenRandom() solution with a BCryptGenRandom() solution.