xLedger: Immutable AI Decision Audit Trail

The Problem

AI systems make high-stakes decisions affecting loans, hiring, medical diagnosis, and criminal justice, but audit trails are easily tampered with or deleted.

Regulators require explainable AI with complete decision records. Traditional database logs can be altered by administrators. There is no cryptographic guarantee that the audit trail reflects reality.

The EU AI Act mandates that high-risk AI systems maintain tamper-evident logs of every decision. Non-compliance carries fines up to 7% of global revenue.

The Old Way

The PRIVATE.ME Solution

xLedger creates an HMAC-chained, DID-signed decision log where every AI decision is cryptographically linked to the previous entry. Tampering with any entry invalidates the chain.

Each decision record includes: input hash, model ID, output, confidence score, timestamp, and the signer’s DID. Records are HMAC-SHA256-chained so any modification invalidates all subsequent entries.

Decision logs can be XorIDA-split across independent custodians for additional tamper resistance. Verification requires only the chain of HMAC hashes, not the raw data.

The New Way

How It Works

xLedger provides an append-only decision log with HMAC-SHA256 chaining, DID signatures, and optional XorIDA split-storage.

HMAC-Chained Hash-Linked Audit Trail

Each decision entry in xLedger is cryptographically linked to the previous entry using HMAC-SHA256, creating an immutable chain where any tampering is immediately detectable.

Chain Structure: Each entry contains the HMAC hash of the previous entry. Entry N includes HMAC(entryN-1) in its data, then computes HMAC(entryN) which is stored in Entry N+1. This creates a forward-linked chain where modifying any entry invalidates all subsequent entries.

Tamper Evidence: To verify the chain, start from the first entry and compute HMAC hashes forward. If any entry's hash doesn't match the next entry's stored hash, the chain is broken. An attacker who modifies Entry 50 cannot recompute the hashes for Entries 51-1000 without detection, because they don't have access to the HMAC key.

Per-Share Protection: When decision logs are XorIDA-split across multiple custodians, each share maintains its own HMAC chain. Tampering with a single custodian's share breaks that share's chain, making reconstruction impossible even if threshold shares are collected.

Use Cases

Integration

import { DecisionLog } from '@private.me/decisionlog';

const log = DecisionLog.create({ signerDid: modelDid });
await log.record({
  input: inputHash,
  output: prediction,
  confidence: 0.94,
  model: 'risk-model-v3'
});

Security Properties

Verifiable Data Protection

Every operation in this ACI produces a verifiable audit trail via xProve. HMAC-chained integrity proofs let auditors confirm that data was split, stored, and reconstructed correctly — without accessing the data itself.

Ready to deploy xLedger?

Talk to Ren, our AI sales engineer, or book a live demo with our team.

Ship Proofs, Not Source

xLedger generates cryptographic proofs of correct execution without exposing proprietary algorithms. Verify integrity using zero-knowledge proofs — no source code required.

Use Cases