A Testable True Random Number Generator for Linux Security Applications

Abstract

Security application extensively utilize randomly generated numbers. Generating these strictly unpredictable numbers is a challenge. My Master's thesis is about delivering a stable stream of entropy to security applications.

A system on chip design which generates random numbers will need 3 distinct modules: A module as entropy source, a module for post processing and a module for communication with another device. The entropy generated originates from a randomly oscillating signal which is sampled. The sampled data will be whitened, to reduce any hardware bias. Finally, the generated data will be tested for randomness and be prepared for transmission to a receiving device.

The proposed design will be made with an "AVNet S6LX9 MicroBoard" development board. A ring oscillator based design will generate the entropy for the system. Any hardware bias will be reduced from the generated data stream. Finally, a MicroBlaze softcore processor will evaluate the generated data and proceed to transfer the stream if the evaluation is successful.