Skip to main content

Understanding Alerts

Alert Severity Levels

SeverityColorMeaningResponse timeExample
CRITICALRedImmediate action requiredMinutesC2 beacon, data exfiltration
HIGHOrangeInvestigate within 1 hour1 hourPort scan, lateral movement
MEDIUMYellowReview within 24 hours24 hoursShadow AI usage, unusual DNS
LOWGreenInformationalBest effortNew device connection

SHAP Explanations

Every alert includes a SHAP (SHapley Additive exPlanations) waterfall chart showing which network features contributed to the detection and by how much.

Example: A C2_BEACON alert at score 0.87 might show:

  • Non-standard port 4444: +0.42
  • Known malicious IP: +0.38
  • Off-hours connection: +0.27
  • High upload ratio: +0.15
  • Base value: -0.35

This is key for:

  • Analysts: Understand why an alert was triggered, not just that it was
  • Auditors: Verify detection logic is sound and unbiased
  • Compliance: NIS2 Art.23 requires explainable incident detection
  • EU AI Act: Decision Traces provide full algorithmic transparency

Alert Actions

ActionWho can do itWhat it does
Block IPAdmin, AnalystAdds source IP to whitelist/blocklist
Mark False PositiveAdmin, AnalystFlags alert as FP, improves future detection
Generate NIS2 PDFAdmin, Analyst, AuditorCreates NIS2 Art.23 incident report
Generate GDPR PDFAdmin, Analyst, AuditorCreates GDPR Art.33 breach report
Escalate to IRAdmin, AnalystMarks alert for incident response team
View Decision TraceAll rolesOpens full proof trail with inputs, factors, outputs
  • Status: Open, Reviewed, Resolved, False Positive
  • Severity: Critical, High, Medium, Low
  • Rule: Filter by specific detection rule
  • Date range: Custom time window
  • Search: Free-text search across alert fields

NIS2 72-Hour Countdown

For CRITICAL alerts, the dashboard displays a countdown timer for the NIS2 Art.23 reporting deadline (72 hours from detection). This helps ensure your organization meets the mandatory incident notification timeline.

Proof Trail

Each alert has a cryptographic proof trail (JSON) that includes:

  • Raw detection inputs (anonymized flow metadata)
  • NDR scoring factors
  • Rule match details
  • SHAP feature contributions
  • Timestamp chain

This proof trail can be exported and provided to auditors or CSIRT as evidence of detection methodology.