Information-Theoretic Conditioning/Post-Processing in SP 800-90B and AIS 20/31 (N23a)
Entropy sources and random number generators (RNGs) often use deterministic processing to improve the quality of raw noise source bits or raw random numbers. In SP 800-90B, this is referred to as “conditioning,” while AIS 20/31 refers to it as “post-processing.” Both standards differentiate between cryptographic and non-cryptographic methods, requiring evidence that non-cryptographic methods are appropriate for the noise source.
This talk examines how AIS 20/31 and SP 800-90B evaluate post-processing algorithms. It focuses on algorithms like almost-universal hash functions and strong blenders, commonly used in entropy sources based on quantum phenomena. The discussion includes the mathematical proofs underpinning these algorithms and their ability to provide nearly perfect output bits under certain assumptions.