Random Number Generators¶
The base class
RandomNumberGenerator is in the header
The major interfaces are
randomize(uint8_t *output_array, size_t length)¶
Places length random bytes into the provided buffer.
add_entropy(const uint8_t *data, size_t length)¶
Incorporates provided data into the state of the PRNG, if at all possible. This works for most RNG types, including the system and TPM RNGs. But if the RNG doesn’t support this operation, the data is dropped, no error is indicated.
randomize_with_input(uint8_t *data, size_t length, const uint8_t *ad, size_t ad_len)¶
Like randomize, but first incorporates the additional input field into the state of the RNG. The additional input could be anything which parameterizes this request. Not all RNG types accept additional inputs.
randomize_with_ts_input(uint8_t *data, size_t length)¶
Creates a buffer with some timestamp values and calls
Generates a single random byte and returns it. Note that calling this function several times is much slower than calling
randomizeonce to produce multiple bytes at a time.
The following RNG types are included
HMAC DRBG is a random number generator designed by NIST and specified in SP 800-90A. It seems to be the most conservative generator of the NIST approved options.
It can be instantiated with any HMAC but is typically used with SHA-256, SHA-384, or SHA-512, as these are the hash functions approved for this use by NIST.
system_rng.h, objects of
System_RNG reference a single
(process global) reference to the system PRNG (such as
You can also use the function
system_rng() which returns a
reference to the global handle to the system RNG.
AutoSeeded_RNG is type naming a ‘best available’ userspace PRNG. The exact definition of this has changed over time and may change in the future, fortunately there is no compatability concerns when changing any RNG since the only expectation is it produces bits indistinguishable from random.
Note well: like most other classes in Botan, it is not safe to share
an instance of
AutoSeeded_RNG among multiple threads without
The current version uses the HMAC_DRBG with SHA-384 or SHA-256. The initial seed is generated either by the system PRNG (if available) or a default set of entropy sources. These are also used for periodic reseeding of the RNG state.
This is a very fast userspace PRNG based on ChaCha20 and HMAC(SHA-256). The key for ChaCha is derived by hashing entropy inputs with HMAC. Then the ChaCha keystream generator is run, first to generate the new HMAC key (used for any future entropy additions), then the desired RNG outputs.
This RNG composes two primitives thought to be secure (ChaCha and HMAC) in a simple and well studied way (the extract-then-expand paradigm), but is still an ad-hoc and non-standard construction. It is included because it is roughly 20x faster then HMAC_DRBG, and certain applications need access to a very fast RNG.
This RNG type directly calls the x86
rdrand instruction. If the instruction
is not available it will throw at runtime, you can check beforehand by calling
This RNG type allows using the RNG exported from a TPM chip.
This RNG type allows using the RNG exported from a hardware token accessed via PKCS11.
EntropySource is an abstract representation of some method of
gather “real” entropy. This tends to be very system dependent. The
only way you should use an
EntropySource is to pass it to a PRNG
that will extract entropy from it – never use the output directly for
any kind of key or nonce generation!
EntropySource has a pair of functions for getting entropy from
some external source, called
pass a buffer of bytes to be written; the functions then return how
many bytes of entropy were gathered.
Note for writers of
EntropySource subclasses: it isn’t necessary
to use any kind of cryptographic hash on your output. The data
produced by an EntropySource is only used by an application after it
has been hashed by the
RandomNumberGenerator that asked for the
entropy, thus any hashing you do will be wasteful of both CPU cycles
On Unix platforms, the
clone() system calls can
be used to spawn a new child process. Fork safety ensures that the
child process doesn’t see the same output of random bytes as the
parent process. Botan tries to ensure fork safety by feeding the
process ID into the internal state of the random generator and by
automatically reseeding the random generator if the process ID
changed between two requests of random bytes. However, this does
not protect against PID wrap around. The process ID is usually
implemented as a 16 bit integer. In this scenario, a process will
spawn a new child process, which exits the parent process and
spawns a new child process himself. If the PID wrapped around, the
second child process may get assigned the process ID of it’s
grandparent and the fork safety can not be ensured.
Therefore, it is strongly recommended to explicitly reseed any userspace random generators after forking a new process. If this is not possible in your application, prefer using the system PRNG instead.