On the Effectiveness of True Random Number Generators Implemented on FPGAs

Randomness is at the core of many cryptographic implementations. True random number generators provide unpredictable sequences of numbers by exploiting physical phenomena. This work compares multiple literature proposals of true random number generators targeting FPGAs. The considered TRNGs are obtained as the combinations of three digital noise sources, namely, NLFIRO, PLL-TRNG, and ES-TRNG, and three post-processing techniques, namely, XOR, Von Neumann, and LFSR. The resulting combinations of such components are evaluated in terms of security, throughput, and resource utilization. The experimental results, which were collected on Xilinx Artix-7 FPGAs, highlight the importance of the post-processing stage for security purposes and reveal NLFIRO as the best digital noise source and LFSR as the best post-processing technique, having the highest throughput with excellent security performance without compromising area and power consumption.