Increasing Robustness of True Random Number Generators against Attacks
Abstract
Random number generation is a critical issue in numerous cryptographic applications: it is used for generation of initialization vectors, challenges, nonces and confidential keys. A flaw in security of the random number generation impacts directly the security of the whole cryptographic system. Random number generators are classified into two main categories: deterministic random number generators (or deterministic random bit generators) and physical true random number generators (P-TRNG). While the deterministic generators are based on algorithmic processes, the P-TRNGs exploit noisy analog phenomena in electronic devices to produce random bit streams. The vulnerability of the physical true random number generators (P-TRNG) to non-invasive attacks has been recently highlighted. Designers should now take into account such vulnerabilities in order to increase the robustness of their generators against attacks. Common fault mitigation techniques such as error detection/correction using redundancy in hardware, time and data do not work in random number generation. For this reason, we propose another method based on real time entropy estimation using embedded tests of source of randomness. The tests are adapted to the structure of the generator and ensure that the randomness of the generated sequence is due to the desired noisy physical phenomenon and that it is not manipulated.