Extracting Cryptographically Secure Uniform Random Bits from a Biased Sequence

Random Number Generators (RNG) are indispensable in cryptography, network security, and statistical simulations. Most commonly available pseudo-RNG (PRNG) are known to produce non-uniform sequences. Natural and hardware sources of randomness are also prone to bias. Therefore, extracting uniform random bits from non-uniform sources of entropy is frequently necessary. In this paper, we look at the philosophical idea of randomness, the impossibility of determining unbiasedness with certainty, and the bias of natural systems. We then propose different methods of doing this, and evaluate them using the NIST statistical test collection. We find that hashing adequate number of bits from a biased sequence is a statistically robust, effective, fast, and efficient way to generate a uniform sequence.