Terminology Index

Glossary

2953 terms

Open concept maps
S
32

Statistical Detection

A detection-engineering approach that builds rules on statistical properties of data, distributions, thresholds, behavioral baselines, complementing signature- and rule-based methods with behavior-driven coverage.

Statistical detection treats detection as a math problem: model what is normal, alert on what deviates significantly. It uses distributions, sliding-window baselines, z-scores, percentiles, and similar techniques to find behavior unlike historical norms. Compared with signature-based detection (which catches known patterns), statistical detection catches the unfamiliar but requires tuning to avoid false positives. As a detection-engineering topic, it underpins higher-tier detection on the Pyramid of Pain.

Introduced in: Detection Engineering

Examples

  • Building a baseline of normal logon patterns and alerting on deviations.
  • Using z-scores in detections to flag statistically unusual events.
  • Combining signature and statistical detection for layered coverage.