Threat coverage matrix
Adversary-led view: attack vectors × the detections that watch them. An unwatched vector reads exactly like an unmapped requirement — 2 gaps right now.
| Attack vector | Kill-chain stage | Detections | Coverage | Linked risks | Residual at stake |
|---|---|---|---|---|---|
Credential stuffing Breach-corpus replay against login endpoints. FKC-2FKC-3T1110.004 | Credential access | CTL-031Adaptive MFA · TPR 95% · FPR 1% · MTTD 1mCTL-112Credential-stuffing WAF rules · TPR 88% · FPR 2% · MTTD 4mCTL-007Session anomaly detection · TPR 60% · FPR 5% · MTTD 12m | strong | RSK-014 | $5.10M |
AitM / session-token theft Adversary-in-the-middle phishing kits (Evilginx-style) stealing live sessions. FKC-3T1539T1557 | Credential access | CTL-007Session anomaly detection · TPR 45% · FPR 6% · MTTD 38m | partial | RSK-014 | $5.10M |
SMS pumping / AIT Artificially inflated traffic to premium-rate numbers. FKC-5MA-01 | Monetization | — none — | gap | RSK-021 | $2.10M |
Insider exfiltration Bulk export of customer data by privileged users. T1567 | Exfiltration | CTL-054DLP export alerts · TPR 50% · FPR 8% · MTTD 4h | partial | RSK-033 | $980k |
Promo abuse Synthetic accounts farming referral and promo credits. FKC-4 | Monetization | — none — | gap | RSK-052 | $450k |
Threat-framework library
Every technique in the shared libraries, checked against the register. Watched means a vector maps to it and has a detection; modeled means a vector exists but nothing watches it; unmodeled means nobody has decided whether it applies here.
Fraud kill chain
2/6 watched
MITRE ATT&CK
4/7 watched
Messaging abuse
0/4 watched
How this feeds the graph
Detections are controls with sharper metrics — true-positive rate, false-positive rate, mean time to detect — so effectiveness is measured, not asserted. A red cell here raises the blindness factor on every linked risk, which pushes it up the dig-deeper queue on the dashboard.