A “Zero Noise” Approach to More Effective Cloud Detection

Summary
The article outlines a defensive method for reducing cloud alert noise: tailor detections to critical assets, review their effectiveness, and investigate every alert. A financial services organization used these steps to reduce noise and detect later attack attempts.
Key points
- Build tailored alerts around critical assets, environment baselines, and known threats; use red teaming to identify visibility gaps.
- Periodically review each detection’s alert volume, false positives, and triage cost, then improve or remove noisy rules.
- Investigate every alert, including false positives, and address the cause by changing detection logic or organizational practices.
- In the case study, targeted alerts for irregular financial transactions helped detect and block subsequent malicious attempts.
- The organization removed noisy DMZ detections and refined load-balancer rules after identifying common sources of false positives.
- After attackers repeatedly used PsExec for lateral movement, the organization banned it internally, reducing noise and making its use a stronger indicator of attacker activity.
Article Details
- Defense Focus
- Reduce cloud detection noise while preserving high-fidelity detection of attacker activity through tailored baselines, feedback on every detection, and thorough alert triage.
- Detection Methods
- Build attacker-focused detections around mapped critical assets and environment baselines, informed by continuous red teaming.
- Monitor transaction servers for attacker-specific irregularities, including inconsistencies in transaction amounts and timestamps.
- Analyze detections periodically for alert volume, false-positive rates, and triage effort; refine or remove detections that are too noisy or ineffective.
- Investigate every alert, including false positives, and use findings to improve detection logic or organizational practices.
- Use exclusions for known legitimate activity, threshold adjustments, and refined search criteria to reduce false positives.
- Data Sources
- Transaction amounts and timestamps on financial transaction management servers
- Access activity involving load balancers and points of sale
- Detection alert outcomes, including false-positive counts and triage time
- Rule Types
- Tailored behavioral detections
- Transaction anomaly detection
- Access monitoring detections
- Detection rules with legitimate-activity exclusions, adjusted thresholds, and refined search criteria
- Automated triage playbooks
- Defensive Actions
- Map critical assets and crown jewels, and use continuous red teaming to identify visibility gaps.
- Review every detection regularly and remove, improve, or otherwise address detections that are noisy or ineffective.
- Triage every alert and investigate false positives rather than setting them aside.
- Coordinate with fraud teams to create high-priority detections for irregular transactions.
- Incorporate known false-positive activities into automated triage playbooks and exclude them from detection rules where appropriate.
- Consolidate administrative tools, use pre-approved accounts or jump servers, and establish clear IT and dev-ops security practices.
- Ban PsExec for internal use where feasible to reduce benign-use noise and make its use detectable.
