Random number generator (device)

Task

If your system has a means to generate random numbers involving not only a software algorithm (like the /dev/urandom devices in Unix), then:

show how to obtain a random 32-bit number from that mechanism.

Batch File

The dynamic environmental variable %random% contains a number between 0 and 32767. <lang dos> @echo %random% </lang>

BBC BASIC

Works with: BBC BASIC for Windows

Requires Windows XP or later. <lang bbcbasic> SYS "SystemFunction036", ^random%, 4

     PRINT ~random%</lang>

arc4random() appeared in OpenBSD 2.1 and has spread to many BSD systems. This function runs an ARC4 random number generator that takes entropy from a kernel device. (This kernel device is sysctl kern.arandom in OpenBSD, or /dev/urandom in some other systems.)

<lang c>#include <inttypes.h> /* PRIu32 */

include <stdlib.h> /* arc4random */
include <stdio.h> /* printf */

int main() {

 printf("%" PRIu32 "\n", arc4random());
 return 0;

}</lang>

Library: OpenSSL

OpenSSL can generate random numbers. The default generator uses SHA1. For Unix systems, OpenSSL will gather entropy by reading a kernel device like /dev/urandom, or by using EGD, the Entropy Gathering Daemon. For other systems, OpenSSL might use a different source of entropy.

<lang c>#include <inttypes.h>

include <stdio.h>

include <openssl/err.h>
include <openssl/rand.h>

int main() {

 uint32_t v;

 if (RAND_bytes((unsigned char *)&v, sizeof v) == 0) {
   ERR_print_errors_fp(stderr);
   return 1;
 }
 printf("%" PRIu32 "\n", v);
 return 0;

}</lang>

Windows

Works with: MinGW

<lang c>#include <stdio.h> /* printf */

include <windows.h>
include <wincrypt.h> /* CryptAcquireContext, CryptGenRandom */

int main() {

 HCRYPTPROV p;
 ULONG i;

 if (CryptAcquireContext(&p, NULL, NULL,
     PROV_RSA_FULL, CRYPT_VERIFYCONTEXT) == FALSE) {
   fputs("CryptAcquireContext failed.\n", stderr);
   return 1;
 }
 if (CryptGenRandom(p, sizeof i, (BYTE *)&i) == FALSE) {
   fputs("CryptGenRandom failed.\n", stderr);
   return 1;
 }
 printf("%lu\n", i);
 CryptReleaseContext(p, 0);
 return 0;

}</lang>

C++

std::random_device is a uniformly-distributed integer random number generator that produces non-deterministic random numbers.

Note that std::random_device may be implemented in terms of a pseudo-random number engine if a non-deterministic source (e.g. a hardware device) is not available to the implementation.

See the C++ section on Random number generator (included) for the list of pseudo-random number engines available.

Works with: C++11

<lang cpp>#include <iostream>

include <random>

int main() {

   std::random_device rd;
   std::uniform_int_distribution<long> dist; // long is guaranteed to be 32 bits
   
   std::cout << "Random Number: " << dist(rd) << std::endl;

}</lang>

C#

<lang csharp>using System; using System.Security.Cryptography;

private static int GetRandomInt() {

 int result = 0;
 var rng = new RNGCryptoServiceProvider();
 var buffer = new byte[4];

 rng.GetBytes(buffer);
 result = BitConverter.ToInt32(buffer, 0);

 return result;

}</lang>

Park-Miller random number generator <lang csharp> const long m = 2147483647L; const long a = 48271L; const long q = 44488L; const long r = 3399L; static long r_seed = 12345678L;

public static byte gen() {

  long hi = r_seed / q;
  long lo = r_seed - q * hi;
  long t = a * lo - r * hi;
      if (t > 0)
          r_seed = t;
      else
          r_seed = t + m;
      return (byte)r_seed;

}

public static void ParkMiller(byte[] arr) {

  byte[] arr = new byte[10900000];
   for (int i = 0; i < arr.Length; i++)
               {                       
                      arr[i] = gen();
               }

}</lang>

ChucK

<lang c> Math.random2(-(Math.random()),Math.random(); </lang>

Common Lisp

<lang lisp>(defun random-int32 ()

 (with-open-file (s "/dev/random" :element-type '(unsigned-byte 32))
   (read-byte s)))</lang>

D

Example for MersenneTwisterEngine for generating uniformly-distributed 32-bit numbers with a period of 2 to the power of 19937. <lang d> import std.stdio; import std.random;

void main() {

 Mt19937 gen;
 gen.seed(unpredictableSeed);
 auto n = gen.front;
 writeln(n);

} </lang>

Output:

run 1: 3500391376
run 2: 9537841895
run 3: 1588499117
run 4: ...

EchoLisp

No random device provided by the host (browser). But we can use the system timer to get a physical input. <lang lisp> (random-seed "simon") (random (expt 2 32)) → 2275215386 (random-seed "simon") (random (expt 2 32)) → 2275215386 ;; the same

(random-seed (current-time-milliseconds )) (random (expt 2 32)) → 4061857345 (random-seed (current-time-milliseconds )) (random (expt 2 32)) → 1322611152 </lang>

Forth

<lang forth>variable rnd

randoms ( n -- )

 s" /dev/random" r/o open-file throw
 swap 0 do
   dup rnd 1 cells rot read-file throw drop
   rnd @ .
 loop
 close-file throw ;</lang>

Fortran

Using system /dev/urandom in GNU/Linux.

<lang fortran> !----------------------------------------------------------------------- ! Test Linux urandom in Fortran !----------------------------------------------------------------------- program urandom_test

 use iso_c_binding, only : c_long
 implicit none

 character(len=*), parameter :: RANDOM_PATH = "/dev/urandom"
 integer :: funit, ios
 integer(c_long) :: buf

 open(newunit=funit, file=RANDOM_PATH, access="stream", form="UNFORMATTED", &
      iostat=ios, status="old", action="read")
 if ( ios /= 0 ) stop "Error opening file: "//RANDOM_PATH

 read(funit) buf

 close(funit)

 write(*,'(A,I64)') "Integer:     ", buf
 write(*,'(A,B64)') "Binary:      ", buf
 write(*,'(A,Z64)') "Hexadecimal: ", buf

end program urandom_test </lang>

Here's an example of the use of the latter:

FreeBASIC

FreeBASIC can in theory use any C library to produce pseudo-random numbers including those which are partly device based.

However, in practice, there is little need for this as specifying a second parameter of 5 to FB's Randomize statement produces cryptographically strong pseudo-random numbers using either the Win32 Crypto API or the /dev/urandom device under Linux. <lang freebasic>' FB 1.05.0 Win64

Randomize , 5

'generate 10 cryptographic random integers in the range 1 To 100 For i As Integer = 1 To 10

 Print Int(Rnd * 100) + 1

Go

In the Go library is crypto/rand, a source specified to use dev/urandom on Unix-like systems and the CryptGenRandom API on Windows. Also implemented here is a source using dev/random, if you really want it. On my system it would print a few numbers then hang until I moved the mouse or pressed some keys on the keyboard. <lang go>package main

import (

   "crypto/rand"
   "encoding/binary"
   "fmt"
   "io"
   "os"

)

func main() {

   testRandom("crypto/rand", rand.Reader)
   testRandom("dev/random", newDevRandom())

}

func newDevRandom() (f *os.File) {

   var err error
   if f, err = os.Open("/dev/random"); err != nil {
       panic(err)
   }
   return

}

func testRandom(label string, src io.Reader) {

   fmt.Printf("%s:\n", label)
   var r int32
   for i := 0; i < 10; i++ {
       if err := binary.Read(src, binary.LittleEndian, &r); err != nil {
           panic(err)
       }
       fmt.Print(r, " ")
   }
   fmt.Println()

}</lang>

Groovy

Based, necessarily, on Java solution: <lang groovy>def rng = new java.security.SecureRandom()</lang>

Test: <lang groovy>(0..4).each { println rng.nextInt() }</lang>

Output:

380425053
-1003791794
-1972330603
1152610574
714616658

Icon and Unicon

The following is Unicon-specific but trivially converted into Icon.

<lang unicon>procedure main(A)

   n := integer(A[1])|5
   every !n do write(rand(4))

end

procedure rand(n)

   f := open("/dev/urandom") | stop("Cannot get to urandom!")
   x := 0
   every !n do x := x*256 + ord(reads(f,1))
   close(f)
   return x

end</lang>

Sample runs:

J

Untested: <lang j>256#.a.i.1!:11'/dev/urandom';0 4</lang>

Fallback: <lang j>256#.a.i.4{.host'dd if=/dev/urandom bs=4 count=1'</lang>

Note: this assumes that J is running on linux.

Java

<lang java>import java.security.SecureRandom;

public class RandomExample {

 public static void main(String[] args) {
   SecureRandom rng = new SecureRandom();

   /* Prints a random signed 32-bit integer. */
   System.out.println(rng.nextInt());
 }

}</lang>

jq

jq does not provide direct access to /dev/urandom, so in the following we assume the availability of `od`, `tr`, and `fold`, and illustrate how to produce an indefinitely long stream of pseudo-random numbers that are approximately uniformly distributed in the range [0,1].

Assuming the jq program shown below is in a file named uniform.jq, the command-line invocation would be:

od -t x -An /dev/urandom | tr -d " " | fold -w 8 | jq -R -f uniform.jq

<lang jq># allow both upper and lower-case characters def hex2integer:

 explode
 | reverse
 | map(if . > 96  then . - 87 elif . > 64 then . - 55 else . - 48 end)
 | reduce .[] as $c
     # state: [power, ans]
     ([1,0]; (.[0] * 16) as $b | [$b, .[1] + (.[0] * $c)])
 | .[1];

select(length>0) | hex2integer / pow(16;length)</lang>

Notice that the program automatically adjusts the precision based on the length of the hexadecimal numbers presented. Since jq uses IEEE 754 64-bit arithmetic, specifying a larger value to `fold`, such as 10, will produce more precise results.

Julia

Works with: Linux

<lang Julia> const rdev = "/dev/random" rstream = try

   open(rdev, "r")

catch

   false

end

if isa(rstream, IOStream)

   b = readbytes(rstream, 4)
   close(rstream)
   i = reinterpret(Int32, b)[1]
   println("A hardware random number is:  ", i)

else

   println("The hardware random number stream, ", rdev, ", was unavailable.")

end </lang>

Output:

A hardware random number is:  986109744

Haskell

Library: Entropy

Works with: GHC version 7.4.1

<lang haskell>#!/usr/bin/env runhaskell

import System.Entropy import Data.Binary.Get import qualified Data.ByteString.Lazy as B

main = do

 bytes <- getEntropy 4
 print (runGet getWord32be $ B.fromChunks [bytes])</lang>

Kotlin

<lang scala>// version 1.1.2

import java.security.SecureRandom

fun main(args: Array<String>) {

   val rng = SecureRandom()
   val rn1 = rng.nextInt()
   val rn2 = rng.nextInt()
   val newSeed = rn1.toLong() * rn2
   rng.setSeed(newSeed)    // reseed using the previous 2 random numbers
   println(rng.nextInt())  // get random 32-bit number and print it

}</lang>

Lasso

<lang lasso>file(`/dev/urandom`)->readSomeBytes(4)->export32bits</lang>

Output:

723217350

Mathematica

<lang Mathematica>rand32[] := RandomInteger[{-2^31, 2^31 - 1}]</lang>

Example: create array of 10 rand32 numbers <lang Mathematica>Table[rand32[], {i, 1, 10}]</lang>

Output:

{355587317, -869860319, -91421859, 1605907693, 101463390, 891823090, 
-531713717, -1038608428, 1717313407, 674189312}

NetRexx

Works with: Mac OS X

and probably other UNIX systems that provide /dev/random or /dev/urandom random data source devices.

<lang NetRexx>/* NetRexx */ options replace format comments java crossref savelog symbols binary

import java.math.BigInteger

randomDevNameFile = File randomDevNameList = ['/dev/random', '/dev/urandom'] -- list of random data source devices randomDevIStream = InputStream do

 loop dn = 0 to randomDevNameList.length - 1
   randomDevNameFile = File(randomDevNameList[dn])
   if randomDevNameFile.exists() then leave dn -- We're done! Use this device
   randomDevNameFile = null -- ensure we don't use a non-existant device
   end dn
 if randomDevNameFile == null then signal FileNotFoundException('Cannot locate a random data source device on this system')

 -- read 8 bytes from the random data source device, convert it into a BigInteger then display the result
 randomBytes = byte[8]
 randomDevIStream = BufferedInputStream(FileInputStream(randomDevNameFile))
 randomDevIStream.read(randomBytes, 0, randomBytes.length)
 randomDevIStream.close()
 randomNum = BigInteger(randomBytes)
 say Rexx(randomNum.longValue()).right(24) '0x'Rexx(Long.toHexString(randomNum.longValue())).right(16, 0)

catch ex = IOException

 ex.printStackTrace()

end return

/* To run the program in a loop 10 times from a bash shell prompt use: for ((i=0; i<10; ++i)); do java <program_name>; done # Shell loop to run the command 10 times

/

</lang>

Output:

$ for ((i=0; i<10; ++i)); do java RRandomGen; done # Shell loop to run the command 10 times
    -3724652236619320966 0xcc4f60865c70f17a
    -8287324416757903696 0x8cfd8259e0b94eb0
    -2951181559250748016 0xd70b4c02052cfd90
     8171526404483923658 0x716717f863fd3eca
    -4285529734202916706 0xc486bd699676009e
     4783094698411310978 0x4260f74949dc3f82
     6972277496665184225 0x60c28171482d97e1
    -2382194670272317046 0xdef0be919c96f98a
     7952058769071853043 0x6e5b6351938ecdf3
    -1857830580859698636 0xe637a8ee0f000234
$

Nim

<lang nim>var f = open("/dev/urandom") var r: int32 discard f.readBuffer(addr r, 4) close(f) echo r</lang>

OCaml

OCaml's default integers are 31 bits on 32 bits architectures:

<lang ocaml>let input_rand_int ic =

 let i1 = int_of_char (input_char ic)
 and i2 = int_of_char (input_char ic)
 and i3 = int_of_char (input_char ic)
 and i4 = int_of_char (input_char ic) in
 i1 lor (i2 lsl 8) lor (i3 lsl 16) lor (i4 lsl 24)

let () =

 let ic = open_in "/dev/urandom" in
 let ri31 = input_rand_int ic in
 close_in ic;
 Printf.printf "%d\n" ri31;

</lang>

but if we really want 32 bits integers there is a module for this:

<lang ocaml>let input_rand_int32 ic =

 let i1 = Int32.of_int (int_of_char (input_char ic))
 and i2 = Int32.of_int (int_of_char (input_char ic))
 and i3 = Int32.of_int (int_of_char (input_char ic))
 and i4 = Int32.of_int (int_of_char (input_char ic)) in
 let i2 = Int32.shift_left i2 8
 and i3 = Int32.shift_left i3 16
 and i4 = Int32.shift_left i4 24 in
 Int32.logor i1 (Int32.logor i2 (Int32.logor i3 i4))

let () =

 let ic = open_in "/dev/urandom" in
 let ri32 = input_rand_int32 ic in
 close_in ic;
 Printf.printf "%ld\n" ri32;

</lang>

PARI/GP

It works on systems having /dev/urandom and Linux.

<lang parigp>rnd(n=10)=extern("cat /dev/urandom|tr -dc '[:digit:]'|fold -w"n"|head -1")</lang>

The code above creates a new function rnd() which returns cryptographically strong integers with max. 10 random digits from /dev/urandom. rnd(n) returns integer with max. n random digits. No leading zeros.

Output:

rnd() = 3055652197
rnd(20) = 75735303746547944580
...

Pascal

This works with FreePascal on "unixoids": <lang pascal>program RandomNumberDevice; var

 byteFile: file of byte;
 randomByte: byte;

begin

 assign(byteFile, '/dev/urandom');
 reset (byteFile);
 read  (byteFile, randomByte);
 close (byteFile);
 writeln('The random byte is: ', randomByte);

end. </lang>

Output:

>: ./RandomNumberDevice
The random byte is: 9
>: ./RandomNumberDevice
The random byte is: 237

Perl

Typically one would use a module as they will work on UNIX, Win32, and other O/S's. Crypt::Random::Seed, for instance, will use Win32 sources, EGD/PRNGD, /dev/u?random, or if none of those exist for some reason, a userspace entropy method. <lang Perl>use Crypt::Random::Seed; my $source = Crypt::Random::Seed->new( NonBlocking => 1 ); # Allow non-blocking sources like /dev/urandom print "$_\n" for $source->random_values(10); # A method returning an array of 32-bit values</lang> or (similar but many more dependencies): <lang Perl>use Crypt::Random::Source qw/get_weak/; # Alternately get_strong print unpack('L*',get_weak(4)), "\n" for 1..10;</lang>

Or we can read values from /dev/urandom ourselves: <lang Perl>sub read_random {

       my $device = '/dev/urandom';
       open my $in, "<:raw", $device   # :raw because it's not unicode string
               or die "Can't open $device: $!";

       sysread $in, my $rand, 4 * shift;
       unpack('L*', $rand);

}

print "$_\n" for read_random(10);</lang> Whether /dev/urandom is good enough for cryptographic work is debated, though on most UNIX systems it is at least as good as the Win32 Crypto API.

Perl 6

Works with: rakudo version 2016-11

A lazy list of random numbers:

<lang perl6>use experimental :pack; my $UR = open("/dev/urandom", :bin) or die "Can't open /dev/urandom: $!"; my @random-spigot = $UR.read(1024).unpack("L*") ... *;

.say for @random-spigot[^10];</lang>

Output:

PicoLisp

<lang PicoLisp>: (in "/dev/urandom" (rd 4)) -> 2917110327</lang>

PowerShell

<lang PowerShell> function Get-RandomInteger {

   Param
   (
       [Parameter(Mandatory=$false,
                  ValueFromPipeline=$true,
                  ValueFromPipelineByPropertyName=$true, 
                  Position=0)]
       [ValidateScript({$_ -ge 4})]
       [int[]]
       $InputObject = 64
   )

   Begin
   {
       $rng = New-Object -TypeName System.Security.Cryptography.RNGCryptoServiceProvider
   }
   Process
   {
       foreach($count in $InputObject)
       {
           $bytes = New-Object -TypeName Byte[] -Argument $count
           $rng.GetBytes($bytes)
           [System.BitConverter]::ToInt32($bytes,0)
       }
   }
   End 
   {
       Remove-Variable -Name rng -Scope Local
   }

} </lang> <lang PowerShell> 4,8,16,32,64,128 | Get-RandomInteger | Format-Wide {$_} -Column 6 -Force </lang>

Output:

1402572656             432337086              413089699             1404567509            -82797202             -261009960

As hexadecimal: <lang PowerShell> 4,8,16,32,64,128 | Get-RandomInteger | Format-Wide {"0x{0:X}" -f $_} -Column 6 -Force </lang>

Output:

0x24305255             0x916002DD             0x9587046             0x5F236274            0xC0BAF6F0            0xC0B93118

ProDOS

Uses math module: <lang ProDOS>printline -random- </lang>

PureBasic

PureBasic has the source for the random data is the "/dev/urandom" device on Linux or Mac OSX and the "Microsoft Cryptography API" on Windows. <lang PureBasic>If OpenCryptRandom()

 MyRandom = CryptRandom(#MAXLONG)
 CloseCryptRandom()

EndIf</lang>

Python

<lang Python>import random rand = random.SystemRandom() rand.randint(1,10)</lang>

Racket

lang racket

Assuming a device to provide random bits

(call-with-input-file* "/dev/random"

 (λ(i) (integer-bytes->integer (read-bytes 4 i) #f)))

</lang>

REXX

version 1

The 32-bit random number is unsigned and constructed from two smaller 16-bit numbers, and it's expressed in decimal.

Note: the REXX random BIF has a maximum range of 100,000. <lang rexx>/*REXX program generates and displays a random 32-bit number using the RANDOM BIF.*/ numeric digits 10 /*ensure REXX has enough decimal digits*/ _=2**16 /*a handy─dandy constant to have around*/ r#= random(0, _-1) * _ + random(0, _-1) /*generate an unsigned 32-bit random #.*/ say r# /*stick a fork in it, we're all done. */</lang>

Output:

4294967296

version 2

This program generates a random 4 byte character string in the range '00000000'x to 'ffffffff'x <lang rexx>left=0 rite=0 lo=hex(left)hex(rite) Say 'low ' c2x(lo) left=random(0,2**16-1) rite=random(0,2**16-1) rand=hex(left)hex(rite) Say 'random' c2x(rand) left=2**16-1 rite=2**16-1 hi=hex(left)hex(rite) Say 'high ' c2x(hi) Exit hex: Return d2c(arg(1),2)</lang>

Output:

low    00000000
random 3E4C3CDE
high   FFFFFFFF

Ring

<lang ring> nr = 10 for i = 1 to nr

   see random(i) + nl

next </lang>

Ruby

Ruby 1.8.7 introduces the 'securerandom' library. For MRI users, this library tries to get random numbers by loading OpenSSL, or opening /dev/urandom, or calling CryptGenRandom.

Works with: Ruby version 1.8.7

<lang Ruby>require 'securerandom' SecureRandom.random_number(1 << 32)</lang>

Rust

'rand' used to be part of Rust standard library but it was extracted as a 'crate' (https://crates.io/crates/rand)

<lang rust> extern crate rand;

use rand::{Rng, OsRng};

fn main() {

   let mut rng = match OsRng::new() {
       Ok(v) => v,
       Err(e) => panic!("Failed to obtain OS RNG: {}", e)
   };

   let rand_num:u32 = rng.next_u32();
   println!("{}",rand_num);

} </lang>

Sidef

<lang ruby>func urandom() {

   const device = %f'/dev/urandom';

   device.open('<:raw', \var fh, \var err) ->
       || die "Can't open `#{device}': #{err}";

   fh.sysread(\var noise, 4);
   'L'.unpack(noise);

}

say urandom(); # sample: 3517432564</lang>

Tcl

<lang tcl>package require Tcl 8.5

Allow override of device name

proc systemRandomInteger Template:Device "/dev/random" {

   set f [open $device "rb"]
   binary scan [read $f 4] "I" x
   close $f
   return $x

}</lang> Usage: <lang tcl>% puts [systemRandomInteger] 636131349</lang>

UNIX Shell

<lang UNIX Shell>od -An -N 4 -t u4 /dev/urandom</lang>

X86 Assembly

Processors supporting the new RDRAND feature can generate a random 32-bit integer in two instructions: <lang x86>L: rdrand eax jnc L</lang> RDRAND reads the CPU's cryptographically-secure hardware random number generator. The loop is needed because RDRAND can occasionally fail to retrieve a value — it sets the carry flag to indicate whether it succeeded.

XPL0

The random number generator is seeded with the 32-bit system timer each time a program starts. From then on, a linear congruential algorithm is used (that passes the Diehard test suite). Since the Ran intrinsic routine returns a signed positive integer (modulo the argument), the value is limited to 31 bits.

zkl

Translation of: C

Linux: <lang zkl>const RANDOM_PATH="/dev/urandom";

fin,buf:=File(RANDOM_PATH,"r"), fin.read(4); fin.close(); // GC would also close the file println(buf.toBigEndian(0,4)); // 4 bytes @ offset 0</lang>

Output:

2152746403

Library: BSD libc

Library: OpenSSL

Windows

Icon and Unicon

version 1

version 2