Generate random numbers from your OS cryptographically secure RNG

{cryptorng} provides access to your operating system’s cryptographically secure pseudorandom number generator (CSPRNG).

System CSPRNGs provide high-quality, uniform random bytes which are secure against cryptographic analysis.

These bytes are suitable for use in other crypographic processes e.g.

• seeding another random number generator
• as keys, seeds, nonces, salts etc for encryption

What sets {cryptorng} apart from other RNGs in R?

• These numbers are as random as you’re going to get on general-purpose PCs which don’t have specialist RNG hardware.
• There’s no seed to set (the OS sets that internally using accumulated system entropy)
  • Depending on your system, the CSPRNG may be continually re-seeded as more entropy becomes available.
• Multiple threads in parallel won’t read the same bytes.

How can I generate random values sampled from a uniform distribution?

library(cryptorng)
# uniform random bytes
rcrypto(16)

#>  [1] 9d 29 5b 9a 4f 7c fd 4a 1e f1 c1 e1 54 34 30 1d

# uniform random bytes as a hexadecimal string
rcrypto(16, type = 'chr')

#> [1] "95b9bbed52bb1178797488ede2905a43"

# random logical values
rcrypto(16, type = 'lgl')

#>  [1] FALSE FALSE FALSE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE  TRUE FALSE
#> [13] FALSE FALSE FALSE FALSE

# uniform random integers
rcrypto(16, type = 'int')

#>  [1] -1462110675  -149265429    24479471 -1108190907   891595010   786375080
#>  [7]  2062005447    56916768  1845302616  -852269330 -2093358023    43116425
#> [13]  1853964883  1208525937 -1385749686   670879908

# uniform random doubles in the range [0, 1)
rcrypto(16, type = 'dbl')

#>  [1] 0.6866093 0.1157316 0.4953703 0.5625702 0.8608402 0.7837645 0.9046331
#>  [8] 0.9356881 0.2529249 0.6325323 0.2894815 0.7857760 0.2388977 0.8496769
#> [15] 0.2457765 0.4280983

Installation

This package can be installed from CRAN

install.packages('cryptorng')

You can install the latest development version from GitHub with:

# install.packages('remotes')
remotes::install_github('coolbutuseless/cryptorng')

Pre-built source/binary versions can also be installed from R-universe

install.packages('cryptorng', repos = c('https://coolbutuseless.r-universe.dev', 'https://cloud.r-project.org'))

Technical bits

Major operating systems (e.g. macOS, Linux, Windows) now come with CSPRNG functionality built-in.
These RNGs (Random Number Generators) are cryptographically secure - meaning that

• bytes generated pass statistical randomness tests
• it is impossible to reconstruct the prior stream of numbers if the current internal state of the RNG is revealed

These are still only pseudo-random number generators though, as the production of numbers is deterministic given the seed. So a high entropy seed which is kept secret is considered standard practice.

For these OS-provided CSPRNGs, the entropy (and the seeding) is provided by the OS itself - using such things as hardware RNGs, timing jitter, network traffic, disk activity. The initial seed is never revealed to the user, and reseeding may take place to ensure that additional entropy is used when available.

The C function for generating random values varies depending on the OS:

OS	CSPRNG	Notes
macOS, *BSD	arc4random_buf()	macOS arc4rand.c
Linux	SYS_getrandom() via syscall()
Windows	BCryptGenRandom()

All of these random number generators are internally seeded by the OS using entropy gathered from multiple sources and use random number algorithms which are considered cryptographically secure.

C code

This package and code is MIT licensed. Please feel free to incorporate/adapt this code into your own project.

The following is a snapshot of the core function which calls the appropriate CSPRNG for your system.

Note: If your system is not supported, please open an issue with information on your systems CSPRNG and/or /dev/random information.

//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// Get random bytes from the system RNG  (C Callable)
//
// @param buf pre-allocated buffer in which to put the random bytes
// @param n number of bytes.  Note: when a system RNG runs out of entropy
//        it may return fewer bytes than expected. This function throws an 
//        error in this situation
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
void rcrypto(void *buf, size_t n) {
  
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  // macOS and BSD support arc4random_buf()
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#if defined(__APPLE__) || defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__)
  // void arc4random_buf(void buf[.n], size_t n);
  arc4random_buf(buf, n); 
  
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  // Windows: use BCryptGenRandom
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#elif defined(_WIN32)  
  // NTSTATUS BCryptGenRandom(
  //     [in, out] BCRYPT_ALG_HANDLE hAlgorithm,
  //     [in, out] PUCHAR            pbBuffer,
  //     [in]      ULONG             cbBuffer,
  //     [in]      ULONG             dwFlags
  // );
  // dwFlags = BCRYPT_USE_SYSTEM_PREFERRED_RNG - Use the system-preferred random 
  // number generator algorithm. The hAlgorithm parameter must be NULL. 
  size_t status = (size_t)BCryptGenRandom( NULL, ( PUCHAR ) buf, n, BCRYPT_USE_SYSTEM_PREFERRED_RNG );
  // Return value is 'NTSTATUS' value. STATUS_SUCCESS = 0.
  if (status != 0) {
    error("cryptorng_windows() error: Status = %zu.\n", status);
  }
  
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  // Linux use 'Sys_getrandom()'
  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#elif defined(__linux__)
  long status = (long)syscall( SYS_getrandom, buf, n, 0 );
  if (status < 0 || status != n) {
    error("cryptorng_linux() error: Status = %zu.\n", status);
  }
  
#else
#error no secrure rcrypto() implemented for this platform
#endif 
}