PRNGs

PRNG stands for pseudo-random number generator. It is an algorithm that produces sequences that resemble randomness but are determined by an initial value, or seed, and thus are reproducible. The output is statistically random-looking but not truly random.

A PRNG maintains internal state and applies a state transition function to produce new values. Most rely on modular arithmetic or bitwise operations. Key characteristics are period (the length before repetition), uniformity, independence, and reproducibility. In practice, non-cryptographic PRNGs favor speed and long periods, while cryptographic uses require unpredictability and forward security.

Common non-cryptographic PRNGs include linear congruential generators (Xn+1 = (aXn + c) mod m), Mersenne Twister, and XorShift families. More recent designs such as PCG and WELL aim for better statistical properties with smaller state. Cryptographically secure PRNGs (CSPRNGs) are built to resist prediction and state compromise, often using cryptographic primitives or DRBG constructions.

Seed choice matters: seeds should be unpredictable or derived from entropy. Many PRNGs support reseeding to refresh state. In software, PRNGs serve simulations, games, randomized algorithms, and procedural content generation; they are inappropriate for true randomness or high-security needs unless specifically designed as CSPRNGs.

Assessment of PRNGs relies on statistical testing and theoretical properties. A good PRNG has a long period, good statistical behavior, and efficient output with modest state requirements.