nx_hmac.nx

hmac.nx -- HMAC-SHA-256 (RFC 2104, FIPS 198-1). license_tier: INDEPENDENT_REDERIVE genealogy_id: international-research-sources/nist/fips_198_1 Keyed-hash MAC built on sha256.nx. Used throughout TLS 1.3 key schedule (RFC 8446 §7.1) as the HKDF primitive, in older TLS as Finished-message signer, and wherever we need a symmetric authenticator without a fresh nonce (unlike Poly1305). Construction (RFC 2104): HMAC(K, M) = H( (K' xor opad) || H( (K' xor ipad) || M ) ) where H = SHA-256, block size B = 64 bytes, and K' = - H(K) if len(K) > B -- pre-hash oversized keys - K || zeros if len(K) < B -- zero-pad short keys - K if len(K) == B ipad = 0x36 repeated; opad = 0x5c repeated. Why not Poly1305 everywhere: - Poly1305 is one-time-key (P4 in poly1305.nx); reusing a key across messages breaks it. HMAC is many-to-one: a long-lived HMAC key authenticates many messages safely. TLS transcripts span the whole handshake so HMAC is the right tool there. Invariants: HM1 Key processing depends only on key LENGTH, not key VALUE: short keys zero-padded, long keys pre-hashed through SHA- 256 (which is constant-time by construction). HM2 XOR loops read every byte of the block regardless of key content; no early exit. HM3 Output is exactly 32 bytes (SHA-256 digest width). References: RFC 2104 (HMAC), FIPS 198-1, NIST SP 800-107. Test vectors from RFC 4231 (HMAC-SHA-256 specifically). nx_safety_envelope: intended_use: "HMAC-SHA-256 -- TLS 1.3 MAC + HKDF foundation + general keyed-hash authentication" sil_target: SIL3 (authentication primitive; MAC