Cribl Detect Concepts
Preview Feature
Cribl is still developing this Preview feature. We don’t recommend using it in a production environment, because the feature might not be fully tested or optimized for performance, and related documentation could be incomplete.
Access requires App Platform, the Cribl Detect feature, and a Role that grants
Apps.Readaccess to the Cribl Detect app.Please continue to submit feedback through normal Cribl support channels, but assistance might be limited while the feature remains in Preview.
Get familiar with the terms and data flow that Cribl Detect uses to detect security threats and surface findings. For alphabetized definitions, see the Cribl Detect Glossary.
Detections, Findings, and Alerts
Cribl Detect uses a short chain of objects. Each step is a different kind of record:
| Object | What it is | What it carries |
|---|---|---|
| Detection | A rule, or related set of rules, that tests data for a condition. Detections run in a detection engine. | Rule logic, severity, and the tags you assign. |
| Finding | A stored result of a detection. | The detection that produced it, the assets and identities involved, recognized MITRE ATT&CK tactics, techniques, and procedures (TTPs), the data source, and the time it was produced. |
| Alert | A group of findings for analysts to inspect. | The related findings, detection context, and event details. |
| Notification | The event sent when an alert is created. Routing uses the Notifications configuration in Cribl Detect. | A URL to the alert, tags, body text, and a timestamp. |
Severity is required metadata on a detection and on the findings and alerts it produces.
Detection Engines
Cribl Detect uses one of two detection engines to determine where and how a detection runs. Cribl Detect assigns the detection engine based on the detection type: single-event or correlation.
For a security analyst, the question is, “Does a detection exist?” not “Which detection engine will run it?” This means you’ll choose a detection type when you create a custom detection. Cribl Detect uses that type to send the rule to the detection engine that can evaluate it efficiently. When you edit an existing detection, the builder matches the detection engine that already runs it.
Detection Types
- A single-event detection evaluates eligible events in the Cribl Search processing pipeline before storage.
- A correlation detection runs in Cribl Search against lakehouse engine data.
Single-Event Detection
Use a single-event detection when every value needed to decide match or no-match is on a single event. Single-event detections evaluate eligible events in the Cribl Search processing pipeline before storage.
When an event matches, Cribl Stream emits a finding and an OCSF-structured copy of the event. Findings and those events are stored in a lakehouse engine dedicated to Cribl Detect.
For example, if event type is 1234 and severity is critical, emit a finding.
Single-event detections are meant for per-event tests, not multi-event or multi-Dataset logic.
Correlation Detection
Use a correlation detection when the logic needs more than one event, Dataset, transformation, enrichment, thresholds within time windows, or other search-time evaluation. Correlation detections run after storage in Cribl Search against lakehouse engine data.
For example, detect DNS queries whose domain names are within a small Levenshtein distance of known command-and-control hosts.
That’s the path for classic SIEM-style rules that join or window data across sources.
Where Detection Results Are Stored
- Single-event detections evaluate eligible events in the Cribl Search processing pipeline before storage and write findings (and OCSF event copies) to the Cribl Detect lakehouse engine.
- Correlation detections evaluate stored lakehouse data in Cribl Search and produce findings or alerts from that logic.
- Findings remain searchable, so later detections and searches can use them. Search them from Search Security Data with Logs in Cribl Detect.
- Alerts are stored separately from findings, in the
sec_ocsf_detection_alertsDataset. Alert suppression controls how many alerts a burst of findings opens. - Alerts gather the findings and detections that need investigation. Opening an alert is the start of triage, not the end of detection.
For how Cribl Stream, Cribl Search, and lakehouse engines fit together, see Cribl Detect Architecture. For how events are prepared for detection, see About Event Data and Schema Normalization.