What makes an audit log tamper-evident
A log anyone can edit proves nothing. How a hash chain works, what it catches, and what it cannot.
Converge Team, 2026-11-12
A tamper-evident audit log is a record of who did what, built so that changing an entry afterwards can be detected. It does not make tampering impossible. It makes tampering visible, within limits that should be stated as plainly as the feature.
An ordinary log table can be edited by anyone with write access to the database, and nobody would know. An auditor who asks "can you show this record was not altered?" needs more than "we would not do that".
How a hash chain works
Each entry in the log includes a hash of the entry before it. A hash is a short fingerprint of some data: change one character of the data and the fingerprint changes completely.
Because entry 5 contains the fingerprint of entry 4, and entry 6 contains the fingerprint of entry 5, the entries form a chain. If someone edits entry 4, its fingerprint no longer matches what entry 5 holds. If someone deletes entry 4, entry 5 points at something that is not there.
Verification recomputes each entry's hash and its link to the entry before it, and reports the entries that do not match. In Converge a quick check covers the newest 1,000 entries, and a full check walks the whole chain page by page.
What each entry holds
| Field | Why it is there |
|---|---|
| Actor | Who did it |
| Roles | What they were allowed to do at that moment |
| Operation | What was attempted |
| Resource | What it was attempted on |
| Decision | Allowed or denied |
| Details | Enough context to understand the entry on its own |
The entries cover operations that change or expose data, such as applied actions, approvals, queries against sources, exports and write-backs to source systems. Query text is truncated, so the log does not become a second copy of the data it describes.
Redaction without losing the proof
Privacy law can require that data be removed, and a log is not exempt. The difficulty is that removing data from an entry changes the entry, and a changed entry breaks the chain.
Converge handles this by hashing a digest of the details, not the details themselves. Redaction clears the details and keeps the digest. Verification can still recompute the entry's hash, so the actor, roles, operation, resource and decision of a redacted entry remain checked. Each redaction is itself an entry in the log.
Redaction clears the details only. The actor and the resource stay in the entry.
Anchors and the end of the chain
A chain has one weak point by construction: its end. Removing the newest entries leaves nothing after them to disagree. The answer is an anchor: a record, kept apart from the entries, of what the chain looked like at a point in time. Verification in Converge compares the chain with its anchor records and reports an anchored entry that has gone missing or no longer matches.
What a hash chain does not do
This is the part most descriptions leave out.
- It does not stop a determined database administrator. The hashes use no secret key. Someone who can write to the database can rewrite an entry and recompute every hash after it. The anchor records live in the same database, so the same person can change those too.
- It does not protect entries written after the last anchor. They can be removed without a report.
- It does not prove an entry was true when written, only that it has not changed since.
- It does not confirm that a redaction was legitimate. Verification checks a redacted entry's fields; it does not check that a matching redaction was recorded.
- It does not replace access control. Who may read the log, and who may write to the database underneath it, are separate permissions.
What it does give is a clear raising of the bar: a casual or partial edit is caught, and an undetected change requires rewriting the chain from that point forward and the anchor records with it.
Questions to ask
- What exactly does verification recompute, and can it cover the whole log?
- Are redacted entries still verified, or skipped?
- Where are the anchors kept, and who can write to that place?
- Who has write access to the database under the log?
- Does the vendor say what the mechanism cannot detect?
Read more
- Audit logging: what is recorded, how to verify, and the full list of limits.
- Human approval: the approvals the log records.
- Sign-in and roles: the roles each entry names.