AnyTRNG: Generic, High-Throughput, Low-Area True Random Number Generator Based on Synchronous Edge Sampling

In this paper, we present a generic, high-throughput, and low-area true random number generator (TRNG) architecture based on synchronous edge sampling. Our approach exploits the entropy source from the jitter introduced by the clock generator with high precision sampling. The substantial gain in throughput is mainly contributed by the novel idea of using a synchronous tapped delay, ensuring the sampling is performed precisely on the signal edge. Furthermore, the jitter source can be utilized from any clock source, making our architecture generic and compatible with cloud-based field-programmable gate array (FPGA), which prohibits combinatorial loops circuits conventionally used to build a digital oscillator. We evaluate our design by implementing it on the Xilinx Artix-7 FPGA, which yields a throughput of 100 Mbps and requires only four look-up tables (LUTs) and seven flip-flops (FFs). Lastly, our proposed architecture, namely the AnyTRNG, is further validated using the NIST 800-90B entropy assessment, as well as NIST 800-22 statistical test and Dieharder, which achieves an approximate min-entropy of 0.9996 per bit.