CC7.2 · D5 · f1
evidence/monitoring-policy.md — Helix Data Systems security monitoring and alerting policyevidence/alert-summary-q4-2025.csv — Monthly alert summary with triage statistics for Q4 2025evidence/monitoring-coverage-review-q4.json — Quarterly monitoring coverage review with gap analysisevidence/penetration-test-summary.md — Annual penetration test executive summary [noise]| ID | Type | Severity | Finding |
|---|---|---|---|
| F-001 | gap | high | Two production services deployed without monitoring for 25+ days helix-ml-inference was deployed November 20 with partial monitoring (APM only, no SIEM, no alerts). helix-data-lake-ingestion was deployed October 5 with only CloudWatch basic metrics — no Datadog, no... |
| F-002 | gap | high | S3 bulk access alert has persistent false positive rate exceeding 30% threshold The S3 bulk access alert has a false positive rate of approximately 95-97% across all three months (152/156 in Oct, 86/89 in Nov, 135/142 in Dec). Policy Section 3.3 requires tuning within 2 weeks whe... |
| F-003 | gap | medium | Failed login brute force alert acknowledged within SLA only 94-99% of the time The brute force alert is High severity (30-minute SLA, 24/7). In October, 839/847 (99%) were acknowledged within SLA. In November, 308/312 (99%). In December, 285/298 (96%). The December spike (coordi... |
| F-004 | gap | medium | SQL injection alert rule disabled for entire observation period The SQL injection detection rule was disabled on November 1 and remained disabled through the end of the observation period. CISO approval was obtained (TUNE-203) with a documented compensating contro... |
| F-005 | gap | medium | December impossible travel alerts — increasing unresolved investigations Impossible travel alerts in December: 31 total, 4 true positives, 22 false positives, 5 'requires investigation.' Only 28/31 (90%) were acknowledged within SLA. The 5 unresolved investigations are a c... |
| Model | Provider | Score | Recall | Prec. | F1 | Gaps | Reported |
|---|---|---|---|---|---|---|---|
| Sonnet 4.6 | Anthropic | 67% | 100% | 50% | 67% | 5/5 | 10 |
| Opus 4.7 | Anthropic | 91% | 100% | 83% | 91% | 5/5 | 6 |
| GPT-5.5 | OpenAI | 83% | 100% | 71% | 83% | 5/5 | 7 |
| GPT-4.1 | OpenAI | 91% | 100% | 83% | 91% | 5/5 | 6 |
| Haiku 4.5 | Anthropic | 59% | 100% | 42% | 59% | 5/5 | 12 |
| GPT-4o | OpenAI | 43% | 100% | 28% | 43% | 5/5 | 18 |