Design of True Random Number Generator Based on Multi-Stage Feedback Ring Oscillator

True random number generators (TRNGs) play an important role in encryption systems. In this brief, a novel method of generating true random numbers on a field-programmable gate array (FPGA) is proposed based on the random jitter of a multi-stage feedback ring oscillator (MSFRO) as the entropy source. Based on the traditional ring oscillator, a multi-stage feedback structure is added to enlarge the range of clock jitter, and improve the frequency of clock sampling and the randomness of the entropy source. Different from the traditional clock sampling structure, we use the clock jitter signal generated by the MSFRO to sample the clock signal generated by the phase-locked loop (PLL) of the FPGA. The obtained output value is operated by XOR to reduce the deviation of the output value and improve its randomness. The TRNG is implemented in Xilinx Virtex-6 FPGA, which has low hardware resource consumption and high throughput. The entropy source classification, hardware resources and throughput are compared with those of existing TRNGs. The results showed that the proposed TRNG only consumed 24 LUTs and 2 DFFs. Compared with other TRNGs, this design has very low hardware resource consumption and its throughput is up to 290 Mbps. The random bit sequence generated by this TRNG passes the NIST SP800-22 test and the NIST SP800-90B test.

I. Introduction

The true random number generator (TRNG) has an important role in many cryptographic systems, including password generation, authentication protocol, key generation, random padding and digital image encryption [1], [2]. In addition, true random numbers have important applications in numerical calculations, statistical simulations, random sampling and quantum key distribution. The performance metrics of TRNGs include throughput, hardware resource consumption, and random number statistics. TRNGs strictly meet the statistical requirements, and unpredictability, and use random physical processes as entropy sources to generate random numbers. Entropy sources include thermal noise, the metastable state [3], clock jitter [4], chaos [5], and Magnetic Tunnel Junction (MTJ) [6], [7], [8]. If the randomness of the original random bitstream is not good, post-processing operations, such as Von Neumann correction or introducing a hash function, are needed to improve the randomness.