Audit logs
The tamper-evident audit log: what is recorded, how to search and export it, and how to verify a sealed export.
The audit log is the platform's permanent record of every significant action. Each entry captures what happened, who triggered it, who or what it affected, and when — along with the originating IP address, approximate location, and user agent. Entries are only ever added, never edited or deleted by hand, and every sealed entry is cryptographically chained and signed so an exported log can be proven untampered by a third party.
You find the log under Audit Trail → Logs in the Admin Console.
What gets recorded
The log covers the full breadth of platform activity. Broadly, entries fall into these areas:
- Authentication — successful and failed logins, single sign-on token exchanges, federated (social/enterprise) logins, silent authentication, step-up and re-authentication, logout and back-channel logout.
- Multi-factor authentication — factor enrollment, verification successes and failures, recovery-code use, factor removal, and admin-initiated MFA resets.
- User lifecycle — account creation, invitation, profile changes, email/phone/username/password changes, account enable/disable/unlock, and deletion. See User lifecycle.
- Authorization — application access granted and revoked, role and group membership changes, organization-admin grants, and access reviews.
- Administration & configuration — changes to applications, directories, connections, network zones, approval workflows, certification workflows, APIs and API keys, organizations, and system settings.
- Security incidents — breached-password attempts, brute-force detection, access from a forbidden region, login from a new device, and rate-limit breaches.
- Device events — registration and removal, commands sent or queued to a device, and connector online/offline transitions.
- Session events — sessions terminated by the system, recording the completed session's total duration.
- Provisioning — outbound SCIM provisioning operations and their outcomes.
- Approvals — approval-workflow creation and lifecycle events, and access-certification campaigns and review decisions.
- Impersonation — when an administrator starts and ends acting on behalf of another user or organization.
For the complete, itemized list with severities, see Audit event types.
The actor model: who did what to whom
Every entry distinguishes the subject (who or what the event is about) from the actor (who triggered it). This is what lets you answer "what happened to user X?" by searching on the subject, while still seeing who acted.
- For self-service actions (a user changing their own password, logging in), the subject and actor are the same person.
- For administrator-on-user actions (an admin assigning a role to someone, disabling an account), the subject is the affected user and the acting administrator is recorded alongside.
- For system or object changes (creating an application, editing a policy), the acting administrator is the entry's user, and the changed object is named in the entry.
- For machine-driven actions over the API, the actor is recorded as a client (machine) rather than a person, together with the channel it came through (Admin Console, API, provisioning, connector, or system).
Each entry's detail view shows the affected object, the changed fields where applicable, the IP/location/user-agent, and the full event context.
Searching and filtering
The log opens with the most recent entries and loads older ones as you scroll. The toolbar above the list is the same one the user and device lists use:
- Filter rules — add one rule per field and combine them: Event type (one or more of the catalog types), Category and Severity (whole groups of types at once), Timestamp (a from/to range with time, in your own time zone; leave one side blank for an open range), Country and City (from the entry's geolocation), IP address, User email, Username, Application and Directory (pick from the list), Organization, and Description. Text rules offer contains, is, starts with and their negations; the list rules offer is any of / is none of.
- Free-text search — match against the description, IP address, the subject user's name/username/email, or the application name. Press Enter to apply.
- Columns — show or hide columns and drag them into order. IP address, location, severity, category, directory and organization are available in addition to the default timestamp, event, application and user.
- Views — the standard views on the left are ready-made filters: Attacks (brute force, lockouts, breached passwords, forbidden region, unusual location, bot detection), Failed logins, Security & Anomaly Detection, System Configuration & Administration and Errors. Save your own combination of rules and columns as a named view, or copy the URL: every filter lives in the address bar.
- Export CSV — downloads the filtered list in the visible column order (share menu), newest first and capped at 50,000 entries; the dialog says so when the view holds more, so narrow the filter to cover everything.
The list also runs live: turn on Live mode and new entries stream in automatically as they happen, and the activity chart refreshes — no page reload needed. This is useful while you reproduce a problem or watch an incident unfold. The chart and the live feed follow the active filter.
Open any entry to see its full detail panel, including the raw event context.
A few worked examples of combining filters:
- All failed sign-ins from one IP in a window — pick the Failed logins view, add a Timestamp rule with the window and an IP address rule.
- Everything from one country — add a Country rule (for example after the protection dashboard's attack map shows an unexpected origin); add Timestamp to narrow it to the day in question.
- All MFA events for a user — add an Event type rule with the multi-factor events and a User email rule.
- All configuration changes by an admin — pick the System Configuration & Administration view and search the admin's name.
- Everything that happened to one device: open the device's History tab. It lists the entries about that device (commands, update runs, approval, pinned users, policy violations), with its own search, type and date filter; the type list shows only the types that occur for the device. It covers entries written since the tab was introduced, and is shown to admins who may read the whole audit log.
Note: One-time secrets that travel in links — password-reset tokens, email-verification codes — are automatically redacted from log entries, so they are never persisted in the audit trail.
Tamper-evident sealing
Beyond being append-only, the audit log is sealed so that tampering can be detected after the fact:
- Hash chain. Each sealed entry is given a sequence number and a hash computed over the entry's content and the previous entry's hash (an HMAC keyed with a per-deployment secret). Because each hash incorporates the one before it, altering, removing, or reordering any entry breaks the chain from that point onward.
- Cryptographic signature. Each sealed entry is additionally signed with an Ed25519 key. The matching public verification key is published, so anyone holding it can verify a signature without access to the platform.
Sealing happens in the background and never blocks or drops a log write.
You do not have to export to spot a problem: once an entry is sealed, its detail view shows the entry's sealing metadata (sequence number, hash, previous hash) together with a hash-valid indicator. A green check means the entry's recomputed hash still matches; a failed check is flagged directly in the console, so a broken chain link is visible in place rather than only at export time.
What this does and does not prove. Sealing lets you — or an external auditor — take an exported batch of log entries and later prove it was not altered, reordered, or truncated after export, and that it was signed by this platform. It is designed for exporting logs elsewhere and proving integrity later. It is not a claim of non-repudiation against the platform operator.
Verifying an exported log
To independently verify a log export:
-
Export the entries. Retrieve the log entries (via the platform's log API, which includes the sealing metadata — sequence number, previous hash, hash, and signature — alongside each entry's content).
-
Obtain the public verification key. The platform publishes the Ed25519 public key used to sign entries; retrieve it once and keep it on file.
-
Verify. Verification walks the exported entries in sequence order and runs three layered checks:
- a structural pre-scan that confirms the sequence numbers are contiguous and each entry's recorded previous-hash matches the entry before it (this needs no secret);
- a hash recomputation that re-derives each entry's chain hash from its content and compares it to the stored value;
- a signature check that verifies each entry's Ed25519 signature against the published public key.
If the entries were altered, reordered, or a tail was cut off, one of these checks fails and pinpoints where.
The Admin Console surfaces the published public key and the current chain tip so you can confirm an export against the live log.
Per-object change history
Separate from the central audit log, some records keep their own versioned change history — a field-level before/after of how that one object evolved. On a user's detail page, open View history to see each version with the fields that were added, removed, or changed (old value to new value) and when.
Use this when your question is "how did this one record get to its current state?" — it reads as a clean diff timeline for that object. Use the central audit log when your question is "who did what, across everything, in this window?". The two complement each other: the audit log is the cross-cutting record; the change history is the per-object story.
Retention
How long entries are kept depends on your subscription tier:
| Tier | Audit-log retention |
|---|---|
| Trial | 7 days |
| Essential | 30 days |
| Advanced | 90 days |
| Pro | up to 365 days |
Older entries are removed automatically once they pass the retention window. If you need to keep records longer than your tier retains them, export them periodically — exported, sealed batches remain verifiable indefinitely. See Plans & add-ons for the full matrix.
Export
Entries can be exported through the platform's log API for archival, ingestion into a downstream system, or independent verification. The export includes each entry's content and its sealing metadata, which is what makes a later integrity check possible.
Export is on-demand — you pull entries when you need them. There is no live streaming target: the platform does not push a continuous feed to an external log collector. If you keep records in an external system, schedule a periodic export, and always export before entries age out of your tier's retention window (see Retention above) — once removed, they are gone from the live log, though any sealed batch you already exported stays verifiable.
Related
- Audit event types — the full catalog of recorded events.
- Issues — recurring system anomalies, tracked separately.
- User lifecycle — the user state changes that appear in the log.