PRNGs for Masking Applications and Their Mapping to Evolvable Hardware

This paper proposes the use of evolutionary computation for the design and optimization of lightweight Pseudo Random Number Generators (PRNGs). In this work, we focus on PRNGs that are suitable for generating masks and secret shares. Such generators should be lightweight and have a high throughput with good statistical properties. As a proof-of-concept, we present three novel hardware architectures that have an increasing level of prediction resistance and an increasing level of reconfigurability at run-time. We evaluate the three architectures on Zynq, Virtex-6, and ASIC platforms and compare the occupied resources and the throughput of the obtained designs. Finally, we use the Spartan-6 platform for the evaluation of the masked implementation where the masks are obtained via our PRNG.