Secure Hash Algorithm (SHA-256)

A cryptographic data structure that uses hash functions to efficiently verify the integrity of large datasets. Data blocks are hashed into leaf nodes, which are then paired and hashed repeatedly until a single Merkle root is produced. This root acts as a compact, verifiable fingerprint of the entire dataset. In AI audit trails, Merkle trees allow an auditor to prove a specific log entry exists within a massive, immutable record without downloading the entire log—a process called proof of inclusion.

A sequential application of a cryptographic hash function where each link incorporates the hash of the previous entry. The structure is defined as H_n = SHA-256(H_{n-1} || data_n). This creates a tamper-evident sequence: altering any historical record breaks all subsequent hashes. Hash chains are the foundational mechanism for append-only logs, ensuring that once an AI decision is recorded, its position in the chronological sequence is cryptographically locked and any retroactive modification is immediately detectable.

A mechanism for message authentication using SHA-256 in combination with a secret cryptographic key. Unlike a plain hash, HMAC-SHA256 proves both integrity (the data hasn't changed) and authenticity (the data came from a holder of the secret key). In AI audit systems, HMACs are used to sign log entries, ensuring that only authorized system components can write to the audit trail and that entries cannot be forged by an attacker who compromises the storage layer.

A cryptographic mechanism using asymmetric cryptography (e.g., ECDSA, Ed25519) to prove authenticity and integrity. The signer uses a private key to generate a signature over the SHA-256 hash of a document; anyone with the corresponding public key can verify it. For AI audit trails, digital signatures provide non-repudiation—the entity that logged a model inference or governance decision cannot later deny having done so, establishing legal accountability.

The process of embedding a cryptographic hash of an audit log into a public blockchain transaction. By writing the SHA-256 digest of a batch of AI audit records onto a decentralized, immutable ledger like Ethereum or Bitcoin, organizations create an independent, external integrity proof. Even if the local audit database is fully compromised, the anchored hash proves the exact state of the log at that block height, leveraging the economic security of the blockchain for compliance verification.

A storage architecture where data is retrieved by its cryptographic hash rather than a physical location. The SHA-256 digest of a file becomes its permanent address: address = SHA-256(content). For immutable AI audit trails, CAS guarantees data integrity by design—any alteration to a stored log entry changes its hash and thus its address. This also provides automatic deduplication, as identical log entries produce identical addresses, optimizing storage for large-scale compliance archives.