A Novel Random Number Generation Method Using NLP and Cryptographic Techniques: A Pipeline Approach with Word2Vec, Enhanced Hash, and Secure Random

Random number generation (RNG) is essential in cryptography, simulations, statistical sampling, and randomized computing. This study proposes the Pipelined Word2Vec Enhanced Hash Secure Random (PWER) method, which integrates Word2Vec-based semantic seed generation, enhanced SHA-256 hashing using timestamp and salt, and HMAC-DRBG-based secure random generation. The proposed framework was evaluated through a complete seven-configuration ablation study covering C1, C2, C3, C1[Formula: see text]C2, C1[Formula: see text]C3, C2[Formula: see text]C3, and C1[Formula: see text]C2[Formula: see text]C3. Each configuration was tested over ten independent trials with one million generated bits per trial. Quantitative comparisons were also performed with Python MT19937, NumPy PCG64, SHA-256 Counter, os.urandom, and secrets. Randomness was assessed using binary entropy, balance deviation, monobit frequency, block frequency, runs, Kolmogorov–Smirnov statistic, autocorrelation, and the official NIST SP 800-22 Rev. 1a Statistical Test Suite. PWER exhibited near-maximal binary entropy and balanced bit generation, and it satisfied the required pass-proportion criteria across all applicable NIST tests and subtests. Statistical analysis further showed that PWER achieved randomness characteristics comparable to those of established software and operating-system-based generators. These findings demonstrate the potential of integrating semantic and cryptographic components within a unified RNG framework.