CSPRNGs

A cryptographically secure pseudorandom number generator (CSPRNG) is a pseudorandom number generator designed to withstand cryptographic analysis and attack. Its outputs should be indistinguishable from true randomness and unpredictable to any efficient adversary, even if parts of the internal state are known. CSPRNGs are used to generate keys, nonces, initialization vectors, and other values whose predictability could compromise security. They differ from ordinary PRNGs in that their security relies on cryptographic hardness rather than statistical properties alone.

CSPRNGs typically combine a cryptographic primitive with an entropy source. They require fresh entropy input through reseeding to maintain security. Common constructions include DRBGs defined in NIST SP 800-90A—hash-based DRBG, HMAC-DRBG, and CTR-DRBG. Fortuna and Yarrow are historical designs that influenced modern implementations. Implementations often rely on operating system entropy pools or hardware random number generators, and may perform health checks or entropy estimation during reseeding.

Security properties of CSPRNGs include unpredictability and forward secrecy, meaning past outputs remain secure or unrecoverable if the internal state is later compromised. They are designed to be resilient against state compromise extensions, where an attacker who gains state access should not easily derive prior outputs. Practical concerns include the quality of entropy, the frequency of reseeding, and protection against flawed implementations or biased inputs.

Examples and standards: widely used systems include DRBGs in OpenSSL, Windows CNG/CAPI RNG, and Linux kernel entropy pools such as /dev/urandom. Guidance and standardization are provided by NIST SP 800-90A/B/C and related FIPS publications.