A Dual Entropy-Based Digital Random Number Generator

This paper introduces the dual-entropy method for oscillator-based digital random number generators (RNG). The standard model for elaborating jitter-based RNG is expanded to account for the sampling time uncertainty. It is explained that combining the jitter of the sampling clock with the uncertainty of the sampled signal proves to be the most efficient method for harvesting entropy from jitter. Dual-entropy configuration also improves the robustness of the RNG against correlation, interference, and injection-based attacks in hardware. Numerical analysis and experimental results show that the negative effect of external interference on the entropy of the output bitstream is significantly reduced in dual-entropy-based RNG. The proposed method is demonstrated on a ring-oscillator-based design and implemented on FPGA. It is tested using NIST 800-22, Dieharder, and TestU01 batteries and observed to produce successful random bitstreams at a maximum center frequency of 45 MHz on Zedboard Zynq and 114 MHz on ZCU102 Zynq Ultrascale+ evaluation kits without post-processing.