Improving Ring-oscillator-based True Random Number Generators using Multiple Sampling

Abstract

A ring-oscillator-based true random number generator (TRNG) can be implemented using only digital standard cells. However, this requires significant hardware resources to compensate for the low bit rate. In this letter, we propose an improved Fibonacci and Galois ring oscillator (FIGARO) TRNG based on a multiple-sampling technique. We implemented FIGARO TRNGs with and without multiple sampling in the same field-programmable gate array and tested the generators' randomness using the National Institute of Standards and Technology (NIST) random test suite. Our experimental results show that the proposed FIGARO TRNG with multiple sampling requires 3.67-4.76 times fewer resources than when only FIGAROs are used for the same bit rates.