SQL Server ledger verification needs independent digest custody
SQL Server ledger tables can make database changes tamper evident. The feature records changes in a ledger structure and supports verification against database digests. That evidence becomes useful only when someone runs the verification, preserves the expected digest independently, and treats an unexpected result as an incident instead of a routine maintenance warning.
Separate evidence from the protected system
Microsoft's ledger overview describes periodic digest generation and storage outside the database in tamper-proof storage, such as an immutable cloud container or a WORM device. Keeping the only copy beside the database does not preserve evidence if an attacker can alter both.
The operational design should answer four questions:
- Which database or ledger table does the digest cover?
- Where is the independent copy stored, and who can delete or replace it?
- How often is a new digest generated?
- Who reviews a failed verification, and what is preserved before recovery?
The database verification procedure can be resource intensive because it recomputes hashes for ledger transactions. Schedule it against the expected workload in a maintenance window or on a readable copy where the procedure supports that choice. A verification that never completes under production load does not provide timely evidence.
Restore is part of the trust path
A restored database can be functional while having a different ledger history. Test restore and verification together. Confirm how the independent digest storage is selected, whether the restore tool preserves the ledger data, and whether the verification job can reach the evidence store after a disaster.
Do not automatically repair a ledger mismatch during a normal restore. Preserve the database, the verification output, the digest identifier, the restore timestamp, and the operator identity. Microsoft provides a recovery procedure for tampering, but recovery is a response decision that changes the evidence state. It should be documented and approved.
Build a custody record
The custody record should include the database name, ledger version, digest storage location, retention policy, hash or object identifier, generation time, verification time, and result. Keep it outside the database and restrict write access. A read-only monitoring account can consume the result without gaining the ability to erase the evidence.
Alerts should cover a failed verification, a missing digest for the expected interval, an expired retention object, and a change in the verification job identity. A successful job that never reaches the digest store is not a successful evidence run.
Ledger does not replace access control, backup encryption, or transaction auditing. It gives the organization a way to detect specific integrity failures when the evidence path is independent. The review should exercise that property and record the actual verification result.
Top comments (0)