DLP and UEBA Are Asking the Wrong Insider Risk Questions

Source: Cloud Security Alliance

Author: unknown

URL: https://cloudsecurityalliance.org/articles/dlp-and-ueba-are-asking-the-wrong-insider-risk-questions

https://cloudsecurityalliance.org/articles/dlp-and-ueba-are-asking-the-wrong-insider-risk-questions

ONE SENTENCE SUMMARY:

Insider risk persists because DLP and UEBA detect discrete events, not contextual action sequences requiring investigations into intent and change.

MAIN POINTS:

  1. Malicious-insider breaches average $4.99M and take 260 days to identify and contain.
  2. Leaders agree insider risk matters, yet struggle describing effective, proactive insider-risk programs.
  3. Authorized, policy-compliant actions become risky only when viewed as a sequence over time.
  4. DLP is effective for compliance exits, but wasn’t designed to model insider intent.
  5. Rules can’t capture intent, sanctioned-tool abuse, temporal arcs, or non-data-moving misuse.
  6. DLP depends on correct labeling; retyped, pasted, screenshot, or new data often escapes controls.
  7. Shadow AI usage is surging; DLP logs can’t distinguish benign debugging from codebase exfiltration.
  8. UEBA detects deviations, but legitimate work frequently looks anomalous without business context.
  9. Alert queues drive tuning toward silence; rare-but-dangerous cases get suppressed as outliers.
  10. Both tools assume risks are specifiable מראש; insider risk instead requires situational reasoning.

TAKEAWAYS:

  1. Prioritize sequence-based behavioral timelines over isolated policy violations or anomaly scores.
  2. Treat insider risk as an investigation workflow, not primarily a detection-and-alerting problem.
  3. Add organizational context (role changes, approvals, resignations) to interpret identical telemetry differently.
  4. Address sanctioned-channel misuse and AI-driven data handling as first-class insider-risk vectors.
  5. Measure success by completed, actionable investigations shared across Security/HR/Legal, not alert volume.