Skip to content

Correlation Engine

CommonHuman-Lab edited this page Jun 1, 2026 · 1 revision

Correlation Engine

GloomProxy correlates all offensive-security runtime data into unified attack intelligence. Correlation runs reactively on every new finding/replay/scan-finish event and can be triggered retroactively via the API.

What it detects

Analysis Description
Exact duplicates Same scanner + type + normalized URL — fingerprint collision
Near duplicates Same type + similar URL, ≥65% 3-gram evidence similarity (cross-scanner)
Related findings Different types on the same host within the same scan job
Attack chains Multi-step exploit paths inferred from rule-based pattern matching
Endpoint clusters Findings grouped by (host, path-prefix) with attack surface score
Replay lineage Replay sessions → confirmed findings (verified vs derived)
Auth correlation Auth sessions active during finding discovery on the same host

Attack chain rules

Chain inference applies ordered rules against findings on the same host. Each rule is a sequence of finding types matched by case-insensitive substring (e.g. "auth" matches authentication_bypass, auth_failure, etc.).

Chain Steps Confidence
Full Compromise sql_injection → auth → idor 0.90
Auth Bypass → Privilege Escalation auth → idor 0.85
File Upload → RCE file_upload → rce 0.85
SQLi → Auth Bypass sql_injection → auth 0.80
SSRF → Internal Pivot ssrf → information_disclosure 0.80
IDOR → Data Exfiltration idor → information_disclosure 0.80
XSS → CSRF / Session Hijack xss → csrf 0.75
Path Traversal → Credential Disclosure path_traversal → information_disclosure 0.75
Open Redirect → OAuth Token Steal open_redirect → oauth 0.70
Information Disclosure → Auth Bypass information_disclosure → auth 0.70
XSS → Stored XSS Persistence reflected_xss → stored_xss (same path) 0.70
Injection → Lateral Movement injection → ssrf 0.65

Risk score = geometric mean of step severities (info=0, low=1, medium=2, high=3, critical=4).

Correlation API

Method Path Description
GET /api/correlation/summary High-level counts across all findings
GET /api/correlation/chains Attack chains sorted by risk score desc
GET /api/correlation/clusters Endpoint clusters (host+path grouping)
GET /api/correlation/duplicates Duplicate/near-duplicate finding groups
GET /api/correlation/lineage Replay session → finding lineage
GET /api/correlation/findings/{id} All correlations anchored to a finding
POST /api/correlation/run Trigger full retroactive correlation pass

All responses include _vis hints (layout type, color, shape) for direct use by D3/Cytoscape frontends.

Clone this wiki locally