Random Number Generation from Image Data and LFSR

Abstract

In 2021, Masaoka, Ichikawa, and Fujieda proposed generating unpredictable random numbers by sampling a hardware LFSR (Linear Feedback Shift Register) at uneven intervals, where the source of entropy was the fluctuation of sampling intervals. Then, in 2023, Chiba and Ichikawa generated random numbers that passed the NIST test by utilizing weather data downloaded from the Internet as the source of entropy. This study proposes the use of digital images as the entropy source for random number generation. Our evaluation results indicate that a 64-bit LFSR should be used, and that fluctuations of the sampling intervals should be generated with the least significant bit of each pixel value. The random numbers generated by the proposed method passed both the Diehard and NIST SP 800-22 tests.