GPUs and chaos: a new true random number generator

Abstract

For applications where security and unpredictability is of utmost importance, true random number generators (TRNGs) play a heavy role compared to its pseudo-random counterparts. Most TRNGs obtain randomness from physical phenomena such as radio noise, radioactive decay or thermal noise that are unpredictable. These applications usually require external hardware to extract entropy and convert them into digital signals. This paper introduces a TRNGs that utilizes graphics processing units as the source of entropy. Its unpredictable behavior is harnessed by computing chaotic maps that are highly sensitive to slight changes to their control parameters and have pseudo-random behavior. A simple post-processing function based on modular addition and XOR is then used to achieve an unbiased output. The security of the proposed TRNG is evaluated using statistical test suites such as the NIST SP 800-22, DIEHARD and ENT, as well as entropy analysis to determine unpredictability. Results indicate that the proposed TRNG has strong statistical quality of random numbers and high throughput without the need of external specialized equipment.