ISAAC, the fast CSPRNG, reimplemented in a nicer, documented, modern C11 API

LibISAAC offers the tiny and fast ISAAC cryptographically secure pseudo random number generator (CSPRNG), in its 32-bit and 64-bit version wrapped into a modern, ISO C11, documented API, ready for embedded usage.

About ISAAC

Quoting from its website:

ISAAC (Indirection, Shift, Accumulate, Add, and Count) generates 32-bit random numbers. Averaged out, it requires 18.75 machine cycles to generate each 32-bit value. Cycles are guaranteed to be at least 240 values long, and they are 28295 values long on average. The results are uniformly distributed, unbiased, and unpredictable unless you know the seed. [...] ISAAC-64 generates a different sequence than ISAAC, but it uses the same principles. It uses 64-bit arithmetic. It generates a 64-bit result every 19 instructions. All cycles are at least 272 values, and the average cycle length is 216583.

ISAAC and its original source code (on which this version is based on) were created by Bob Jenkins and released into the public domain.

Usage example

// Init the context with a secret seed
isaac_ctx_t ctx;
const uint8_t seed[16] = {1, 2, 3, 4, 5, 6, 7, 8,
                          9, 10, 11, 12, 13, 14, 15, 16};
isaac_init(&ctx, seed, sizeof(seed));

// If you DON'T need ISAAC for security purposes, an all-zero
// seed may also be of interested. Just pass a NULL seed.
isaac_init(&ctx, NULL, 0);

// Extract ANY amount of pseudo-random integers
isaac_uint_t stream[300] = {0};
isaac_stream(&ctx, stream, 300);
isaac_uint_t another_stream[5000] = {0};
isaac_stream(&ctx, stream, 5000);

// Note that ISAAC provides integers, not bytes.
// If you need bytes, convert the integers to little endian-
// or big endian-encoded stream of bytes.
uint8_t bytes[300 * sizeof(isaac_uint_t)];
isaac_to_big_endian(bytes, stream, 300);
// or isaac_to_little_endian(bytes, stream, 300);

// Done using ISAAC? Cleanup the context to avoid leaving traces
// of the state and of the seed.
isaac_cleanup(&ctx);

Include it in your project

Bitness

Depending on your architecture, you may want to use the version of ISAAC optimised for 32 or 64 bit words. To do so, set the #define ISAAC_BITS to 32 or 64 during compilation or directly in the header file. It defaults to 64 when unspecified.

Differences:

the output differs between the two (it's expected to differ)
the context grows twice in size when using the 64 bit version
- 32 bit: 2064 B
- 64 bit: 4128 B
in the 64 bit version, twice the amount of bytes is provided per reshuffling of the state

Please note that altering ISAAC_BITS does not set your compiler to compile LibISAAC for a 32 or 64 target architecture, it simply chooses which integers to use based on your choice, not the compiler's. You may want to choose the best version for your system, but you can compile both versions on any system.

Examples:

a 32-bit microcontroller works best with ISAAC_BITS=32
a 64-bit desktop works best with ISAAC_BITS=64
a 64-bit desktop that has to test the output of ISAAC coming from a 32-bit microcontroller (assuming the same seed is known to both), should compile with ISAAC=BITS=32 to get the same output.

Static source inclusion

Copy the inc/isaac.h and src/isaac.c files into your existing C project, add them to the source folders and compile..

If you prefer a specific bitness, redefine ISAAC_BITS in the header file.

Compiling into all possible targets

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DISAAC_BITS=32
cmake --build .

This will build all targets:

a libisaac32.a static library
a test runner executable testisaac32
the Doxygen documentation (if Doxygen is installed)

To compile with the optimisation for size, use the -DCMAKE_BUILD_TYPE=MinSizeRel flag instead.

If you prefer using 64 bit integers, set -DISAAC_BITS=64.

TheMatjaz / LibISAAC