Optimising ring oscillator-based true random number generators concept on FPGA

Random Number Generators are widely used in cryptographic applications (symmetric and asymmetric encryption, digital certificate generation, etc.) and other related areas. True Random Number Generators (TRNG) are based on physical non-deterministic phenomenon. Most of these generators are based on a large number of interconnected high frequency ring oscillators (RO), which are most commonly implemented in FPGA. This paper describes an optimized solution which increases the speed and the complexity of these ring oscillator-based TRNG. The proposed solution combines a personalized approach on the Ring Oscillator scheme, based on preprocessed raw data and a well-known one, based on the desynchronization of a large number of free running ROs, therefore minimizing the resource used within the FPGA. The paper covers the most known statistical testes (provided by the National Institute of Standards and Technology - NIST) applied on a large amount of data and proves the stability of a generator by providing positive experimental results that were obtained from measurements using different operating frequencies.