FPGA Based True Random Number Generators Using Non-Linear Feedback Ring Oscillators

True random number generators (TRNGs) are important hardware primitives required for many applications including cryptography, communication, and statistical simulation. This paper presents a non-linear feedback ring oscillator (NL-FRO) based entropy source for implementing high performance TRNGs on FPGAs. The proposed NLFRO structures harvest randomness from noise and unpredictable variation in delay cells and bi-stable elements which are further amplified by non-linear feedback loops. The outputs of NLFROs show chaotic behavior, making them suitable for implementing high entropy, high speed and attack resistance TRNGs. Three NLFRO-TRNGs are implemented and tested on an Altera 60nm FPGA device. Raw entropy and statistical properties of the NLFRO-TRNGs are examined by the NIST 800-22 entropy estimation and NIST 800-90B statistical test suits. Compared to the prior art, experimental NLFRO-TRNGs show higher entropy and lower resource usage while consuming sub-milliwatt at 200 Mbps.