Safe cryptographic random number generation using untrusted generators

The security of many cryptographic applications relies heavily on the quality of the random numbers used. Therefore, random number generation is one of the most critical primitives for cryptography. This paper focuses on true random number generators (TRNGs) and the analysis of their security requirements. After illustrating issues associated with adversarial influences on TRNGs, we propose a simple method to obtain a secure TRNG based on n TRNGs originating from (potentially) untrusted vendors. The untrusted generators are combined such that as long as one out of the n vendors does not collude with the other vendors, the generator is secure, i.e., the output is unpredictable and uniformly distributed even in the presence of an active attacker. In order to achieve this, we review several choices of functions to be used as combiner. The advantage of our design is that only the (black-box) input-output behavior of the vendor's TRNGs needs to be evaluated. No overhead is introduced by the combiner. The resulting generator offers faultresilience and ease of maintenance.