Using Evidence and Context to Improve Detection Engineering

· Original article ↗

Summary

Huntress proposes applying criminal-justice concepts to detection engineering, using evidence levels and behavioral context to assess intent, classify alerts, reduce false positives, and prioritize responses.

Key points

  • The article suggests assessing both observable actions and contextual evidence of intent when classifying security alerts.
  • It describes three evidence thresholds—preponderance, clear and convincing, and beyond a reasonable doubt—as analogies for confidence levels and response decisions.
  • Correlating events across data sources, assessing how common behaviors are, and using risk-based alerting can strengthen alert evaluation.
  • Behavioral patterns, user roles, history, and business context can help distinguish malicious activity from legitimate actions.
  • The article recommends considering potential impact alongside evidence when prioritizing alerts and choosing a response.
  • It frames means, motive, and opportunity as additional considerations for evaluating suspicious activity.

Article Details

Defense Focus
Improve alert accuracy by assessing both observable actions and contextual evidence of intent, then calibrate response to confidence and potential impact.
Detection Methods
  • Correlate related events across multiple data sources to strengthen confidence in an alert.
  • Use cardinality and behavioral patterns to identify actions that are rare or unusual in context.
  • Prioritize alerts by potential impact and likelihood of malicious intent, requiring stronger evidence for higher-impact responses.
  • Assess actions against historical behavior, user role, business context, and threat intelligence.
  • Consider means, motive, and opportunity alongside the observed activity when evaluating intent.
Data Sources
  • Logs and correlated events from multiple sources
  • Authentication and login activity
  • File access and transfer activity
  • Privilege and security-control changes
  • Historical behavioral patterns
  • User role and business context
  • Threat intelligence
Defensive Actions
  • Investigate and correlate additional signals before classifying ambiguous alerts as malicious.
  • Prioritize high-confidence, high-impact alerts for immediate response.
  • Require extensive contextual evidence before taking drastic actions such as disabling an account or blocking network traffic.
  • Use business justification and expected user behavior to distinguish benign activity from potential threats.

People