Seed & Hashing: Architecture of Random Outcome Generation

A seed (initial value) is a fundamental parameter that initializes a deterministic pseudorandom number generator (PRNG) algorithm. In cryptographically secure systems, the seed is formed from high-entropy sources: hardware random number generators (HRNG) based on physical processes (thermal noise, shot noise in semiconductors, radioactive decay). The minimum recommended seed length is 256 bits, which provides a state space of 2²⁵⁶ and excludes full brute-force search at the current level of computing technology. The quality of the seed directly determines the cryptographic strength of the entire subsequent sequence.

Hash functions in the outcome generation architecture perform the role of a deterministic transformation of the input seed into an output value of a fixed length. The main requirements for cryptographic hash functions: (1) determinism — the same input always yields the same output; (2) efficiency — calculating the hash takes polynomial time; (3) preimage resistance — it is impossible to recover the input from the output; (4) avalanche effect — changing one bit of the input changes ≥50% of the output bits. In modern systems, SHA-256, SHA-3 (Keccak), and BLAKE2b are used, each of which provides a 128-bit security level against birthday attacks.

Entropy sources are classified into hardware (HRNG) and software (PRNG/CSPRNG). Hardware generators use non-deterministic physical processes and provide true randomness, but are limited in generation speed (usually 10–100 Mbps). Software CSPRNGs (Cryptographically Secure PRNGs), such as ChaCha20, AES-CTR-DRBG, or Fortuna, use a hardware seed for initialization and generate a cryptographically secure sequence with high performance. A hybrid architecture involves periodic updates of the internal state of the CSPRNG with fresh entropy from the HRNG through reseed intervals (typically every 2³² generations).

Verification of the statistical uniformity of the generated sequence is carried out through a battery of standardized tests. The NIST SP 800-22 suite includes 15 tests: frequency test, runs test, binary matrix rank test, spectral test (DFT), Maurer’s universal statistical test, and others. Each test calculates a p-value, and the sequence is considered statistically random if the p-value ≥ α for all tests at a given significance level α (usually α = 0.01). Additionally, the extended suites TestU01 (BigCrush — 106 tests) and PractRand are applied for in-depth analysis of long-period correlations.