Audit log

An append-only record of what happened in DFIRe, who did it, and what changed. Open it from the user menu, under Audit Log.

What an entry holds

  • Action, such as CREATE, UPDATE or DELETE.
  • Record type, the kind of thing the action touched. Cases, evidence items, attachments, notes, timeline events, users, API keys, indicators, settings, backups and webhook deliveries all appear here.
  • Description, a readable summary of the change.
  • Timestamp, to the second.
  • Username of whoever acted.
  • IP address, either the client DFIRe saw directly or the value your trusted reverse proxy sent in X-Real-IP. DFIRe never uses X-Forwarded-For for audit identity, because a proxy that appends leaves attacker-supplied entries in it. Entries written without a request, such as license checks and heartbeats, have no IP address.
  • Change data, a structured before-and-after for every modified field. Empty means that the field had no value. A secret shows as ***. See Secret values.

Some system events carry more in their change data. A heartbeat, for example, records database and Redis latency, queue counts, case and evidence totals, active users, storage use, and the running version and uptime. Together they build a history of the installation's health that you can look back through.

A failed audit write blocks the operation. When DFIRe cannot record an action, it refuses the action rather than completing it unrecorded.

Secret values

An entry never holds a secret. It names the field that changed and shows *** in place of the value. DFIRe treats these as secrets:

  • Credentials in settings, such as passwords, API keys and tokens, and a user's phone number.
  • Any value under a field whose name ends in a word such as password, secret, token or api_key. This includes custom attributes on cases and evidence items. DFIRe records that an attribute named admin_password changed, and leaves out its value.
  • The endpoint URL of a webhook notifier, after its host. The key of such a URL is often in its path or query, so an entry shows https://hooks.slack.com/***.

The field decides, never the text. DFIRe records everything you type into a text field as you typed it, such as a note, a description or a chat message. A cookie or a key in a note can be a live secret or an indicator you keep for threat hunting, and DFIRe cannot tell which. If you write a secret into a note, DFIRe stores it, and the audit log holds it too.

A record always keeps the value of its own fields. DFIRe masks only the copy in the audit log. The container log and forwarding carry that copy.

Refused requests

DFIRe writes a SECURITY_VIOLATION entry for a refused request in the web interface, the REST API, the MCP server, Slack or the Django admin. These reasons count:

  • The user's roles lack the permission.
  • The user's place on the case team does not allow the action.
  • The action is for the case lead only.
  • The request names an existing case that the user may not read.
  • The action is for superusers only.
  • The action needs a signed-in browser session, and the request came through an API key or Slack.
  • A policy of the installation does not allow it. For example, DFIRe does not delete an indicator that a case links to. With owner-only playbook editing, a user cannot change a playbook that another user owns.

The record type of these entries is Permission. The change data names the reason, the operation and the interface: rest, mcp, slack or admin. It also names the case when the case team or the case read refused, and the record when a policy refused. A refusal for a missing permission usually names no case, because DFIRe checks the permission before it looks up the case. It can name the case when the permission depends on what the request does in the case. DFIRe records the same refusal of the same user once per minute, so a client that retries does not fill the log. A refusal for another case or record is a new entry.

DFIRe writes no entry for a refusal that comes from the state of a record, such as a write to a closed case. It also writes none for a refusal caused by another user's change made at the same moment. A live connection of the web interface writes no entry: it closes when its access ends, and it drops a message it may not send.

These entries record where a boundary held, not an attack. Like other account activity, they reach superusers and roles with global case read.

Finding entries

Type in Search entries to search usernames, actions, labels and content. Without the permission to view legal entities, search and the Record filter do not match a legal entity's name in custody, evidence item or project entries. To narrow the list further, select Filter and add a filter row. A row compares one field with a value. User, Record and IP address compare text. Action and Record type take a value from a list, and Time takes a day and an optional time of day. Record type lists the record types you can see in the log.

Rows join with AND, OR or NOT, read from left to right. A NOT row leaves matching entries out: NOT Action equals LICENSE_PING hides the license checks.

Time rows use your display time zone. Without a time of day, a Time row takes the whole day, so Time is after 30 Sep keeps the entries from 1 October on. Add a time of day (24-hour, HH:MM) to narrow the list to the minute. Time is after 30 Sep 14:05 keeps the entries from 14:06. Time is before 30 Sep 14:05 keeps the entries before 14:05. To keep the entries of 14:05 too, use is after 14:04.

Each row shows as a chip under the toolbar. Select a chip to change its row, select its remove control to remove it, or select Clear filters to remove every row. To move a row, drag its chip by the grip at its start. DFIRe reads the rows from left to right, so their order can change the result. The search and the rows are part of the page address, so a reload or a shared link opens the same list. If the list does not load, the page says so instead of showing an empty list.

The audit log shows entries as cards or as a table. Switch between them with Cards and Table above the list, and the browser remembers the choice. The count above the list states how many entries are loaded and how many match the filters, for example 20 of 10,970 entries. Scroll down to load more entries.

The table shows Time, User, Action, Record type, Record and IP address. Open Columns to show, hide or reorder columns. Changes, a one-line list of the changed fields, is off by default. Select the Time, User, Action or Record type heading to sort by that column, and select it again to reverse the order. The table opens newest first, and the cards always show the newest entries first.

Select a row, or the arrow at its start, to open the entry's changes under it. The changes have the same Visual and JSON views as on a card.

Who sees what

Reading the log at all requires the View auditlog permission, granted per role under Settings → Access Roles. Without it a role reaches no audit entries anywhere, including the history shown on an evidence item.

With it, entries are filtered down to what that person could already see.

  • Case records. Entries about a case, its evidence, notes, attachments, timeline, compliance timers, custody transfers, indicators and reports appear only for cases the person can reach. Global read sees every entry in the log, and global edit the entries of the cases it reads.
  • Shared records. Entries about projects, legal entities, runbooks, indicators, playbooks and other shared configuration appear only if the role can read that kind of record. A role that cannot open the Legal Entities directory does not see legal entity history either. It also does not see the legal entities in the entries of projects, evidence items and custody transfers.
  • Their own account. Everyone sees their own sign-ins and account changes.

Other people's account activity, API key use, role changes and system configuration history reach superusers and any role that holds global case read (core.view_all_cases), which opens the log in full. The shipped Team Lead role holds it, so a Team Lead reads every entry in the installation. The configuration history includes the entries about automation rules and webhooks.

A filtered entry is hidden, not missing.

Append-only storage and retention

DFIRe never edits an audit record after writing it, and no screen or endpoint deletes a single record. The application raises an error on any attempt to modify or delete an entry. Old entries leave the log only through retention, which a superuser controls.

What that protection covers. The application enforces it, so it holds against everything reachable through DFIRe except retention, which records itself. It cannot hold against someone with direct access to the database or its backups, because they are past the application entirely.

On a system you host yourself that limit is structural rather than a gap to close with database locks, because whoever runs the installation can change how it works. The way to detect tampering is to keep a copy somewhere they do not control. Forward entries to a separate system as they are written, then compare the two.

Retention

By default DFIRe keeps every entry, and the log grows for the life of the installation. A superuser can limit it under Settings → Log Management, in the Local retention section:

  • Delete old records automatically deletes entries older than a period you set, from 30 to 3650 days. DFIRe applies it once a day.
  • Delete old records now deletes entries older than the number of days you enter. DFIRe shows how many entries that is before it deletes anything.

DFIRe cannot restore a deleted entry. These rules apply to both controls:

  • While forwarding is switched on, DFIRe deletes an entry only after forwarding it. If forwarding is paused or failing, old entries wait until it runs again or you switch it off.
  • Each deletion writes an entry of its own, with who started it, the cutoff date and the number of entries. DFIRe never deletes these entries.
  • DFIRe records a change of the retention period with the old and the new value.

A settings export does not carry the retention period. Set it on each installation.

If you need entries kept for longer, or kept where a DFIRe administrator cannot delete them, forward them.

Forwarding to an external system

DFIRe can send audit entries to a log aggregator or SIEM as they are written. Configure it under Settings → Log Management.

This is also how you make the trail hold up to scrutiny. Keep a copy on a system the DFIRe administrator does not control, then compare the two. A discrepancy shows even when the local record looks consistent.

PropertyDetail
TransportHTTPS POST to an endpoint you choose. DFIRe rejects a plain HTTP URL.
Payload formatJSON Batch for a generic receiver, NDJSON for Logstash or Fluentd, Elasticsearch Bulk for OpenSearch and ELK, or New Relic.
AuthenticationHTTP Basic, or custom headers such as a bearer token. DFIRe encrypts the credentials and headers at rest.
DeliveryCursor-based, so each entry goes exactly once and in order.
BackfillOn first enabling it, choose whether to send the whole history or start from now.
ReliabilityBatch size, timeout and retry count are configurable. A circuit breaker opens after a set number of consecutive failures, so a broken endpoint does not get hammered.
Delivery logRecent attempts stay visible with their timing, response code and any error.

The audit log carries enough detail to satisfy activity-logging and traceability obligations under frameworks such as GDPR and PCI DSS. Forwarding is how you keep that record beyond the DFIRe database.