High-entropy random number generators using system on chip devices

Security-critical applications require "random " numbers for generating cryptographic keys. It is widely known that generating true random numbers will have to address physical phenomena with random proprieties: quantum mechanics or thermal phenomena in electronic devices. More than that, harvesting high-entropy data from hardware generators is usually a difficult task and most of the applications will take data that is believed to contain enough entropy and use it to ldquoseedrdquo a cryptographic algorithm. The paper proposes an evaluation of such system on chip device designed to harvest high-entropy data from thermal phenomena (thermal noise and avalanche noise) and to analyze ldquowhiteningrdquo algorithms for reducing bias and correlation. Taking into account the features of System On Chip devices: complete user management of analog and digital embedded resources, the possibility of creating custom peripherals and the high level of re-usability not only for the software modules but also for embedded hardware peripherals, the new design addresses the main challenges for building true random number generators and proposes a design path and a testing procedure for high-entropy generators.