Anomaly Detection
Flagging activity that deviates from an established normal baseline.
What it is
Anomaly detection identifies activity that statistically deviates from an established baseline of normal behavior, rather than matching a known-bad pattern. It surfaces the unusual, on the theory that attacks often look different from routine operations.
Why it matters
It can catch novel threats that have no signature, but it depends on a solid baseline and tends to be noisier, since unusual is not always malicious.
In practice
Alerting when a service account that normally touches one server suddenly authenticates to fifty.
What to look for
- Spikes in logins, data transfer or process counts against the norm.
- New countries, devices or hours for a user.
- Alerts that cluster around normal business changes.
Common mistake
Treating every anomaly as an attack. Rare is not the same as malicious; add context before acting.
Practice it nearby
CyberKit has no tool for this term itself. These are the closest places to practice it.
- Hunt Query Builder
Build a hunting query in KQL, SPL, LogScale or PowerQuery against documented field names, or start from a prebuilt hunt.
- Rule Range
Write a regex, a Sigma rule, a YARA rule and a query pipeline, each graded live against samples, some of which you never see.
In the field
Well-known tools where this shows up on the job. They are listed to show the landscape, not as a recommendation.
-
Splunk Enterprise Security Commercial
A SIEM where these logs are searched with SPL and turned into correlation searches.
-
Microsoft Sentinel Commercial
Microsoft's cloud SIEM, queried with KQL and fed by connectors for these log sources.
-
Elastic Security Commercial
A SIEM built on Elasticsearch, with a detection rules library and timeline investigations.
-
Wazuh Open source
An open source platform that collects, decodes and alerts on host and log data.
Read next
- Baselining: Establishing what 'normal' looks like so deviations stand out.
- Signature-Based Detection: Detecting known-bad by matching specific, predefined patterns.
- UEBA: Analytics that profile users and entities to flag risky deviations.