Repository navigation
Releases
v0.5.0 — Accurate CVE matching
Accurate CVE matching
SecSuite's network scanner used to report CVEs for the wrong products. v0.5.0 fixes that at the source.
Before: an authorised scan of a production Ubuntu server reported 20 CVE matches, several at CVSS 10.0. None of them applied. They were for Authelia, zzcms, Pulpcore, NGINX Unit, WordPress plugins and UPnP routers, or for nginx and OpenSSH versions fixed years ago.
After: the same scan reports 7 matches, all for the software actually running.
Fixed
- CVE lookup checks affected products and versions. NVD keyword search matches any CVE whose description mentions a product. Each result is now checked against its CPE configuration, NVD's own list of affected products and version ranges, and different-product or out-of-range matches are dropped. For recent CVEs that don't have CPE data yet, a match is kept only when the product is the subject of the description. Version comparison understands patch suffixes such as OpenSSH
9.6p1. - Port-number guesses no longer trigger CVE lookups. When nmap can't fingerprint a service, it names it after the port (5000 →
upnp). Those names are now marked as guessed, skipped for CVE lookup, and labelled in the output.
Added
- Native entity emission for scanners (#14). Scanners can emit their provenance graph directly;
PortScanneris the first to do so.
Upgrading
No configuration changes. Expect fewer, more relevant CVE findings. If you relied on the old keyword results, review your saved scans: many of those findings were false.
Full changelog: v0.4.0...v0.5.0
v0.4.0
Added
- Community-extensible correlation engine (
modules/correlation/). Turns
co-occurring findings into single, higher-signal correlations via declarative
YAML rules (rules/correlation/) instead of hardcoded Python — so detections
can be contributed and reviewed the way Nuclei templates are. Rules are either
flat (predicates matched anywhere in a run) or host-scoped, which groups
findings by the anchor entity they trace back to in the entity graph and emits
the full provenance chain (10.0.0.5 → 445/tcp → CVE-2017-0144). Ships with a
starter pack of five rules (three migrated from the previously hardcoded attack
patterns, plus two provenance-aware ones) and a contributor guide. - Entity-graph builder (
modules/correlation/graph_builder.py) that turns a
run's scan results into a provenance graph (target → host → service/url/ vulnerability), sosecsuite correlatenow activates host-scoped rules
live — findings trace to a shared host anchor and correlations carry the
chain that led to them. Rules can anchor on multiple host types
([ip_address, hostname, domain]). - CVE enrichment in
correlate(modules/correlation/enrich.py). Looks up
CVEs for the services the port scan discovered and nests each under its service
in the graph (host → 445/tcp → CVE-2017-0144), so vulnerability-based rules
likecritical-vuln-on-exposed-hostfire in the CLI path too. Toggle with
--cves/--no-cves. secsuite correlate <target>to run the rules over a scan, and
secsuite rules list/secsuite rules validateto inspect and validate
the rule library (including contributed rules).
v0.3.0
A security + capability release: five real security fixes hardening the scan
inputs and forwarding paths, a large test-coverage jump, and the first piece of
entity-aware analysis — a provenance graph that records how each finding was
reached.
Added
- Entity provenance graph (
core/entities.py). Every discovered thing is now
anEntitylinked to the entity it came from, collected in anEntityGraph
(10.0.0.5 → 445/tcp → CVE-2017-0144), so a report can show how we got here
instead of a flat finding list. Identity is(type, value)with automatic
dedup, cycle-safeancestry(), and sticky in-scope resolution. Prerequisite
for correlation rules and a module registry. (#12)
Security
- Blocked SSRF via OpenAPI spec URLs and spec-supplied servers in apisec —
spec locations and theserversthey declare are now validated before any
request is made. (#5 path) - Required TLS 1.2+ for syslog/SIEM forwarding, so events are no longer sent
over downgradeable transport. - Validated phishing campaign and tracking ids before they are used, closing
an injection path through attacker-influenced identifiers. - Scrubbed untrusted values before logging to prevent log injection, and
repairedCacheEntry.__hash__. - Cleared the CodeQL alerts raised across the new code.
Testing
- Coverage for the exploit/remediation execution paths, the RedBlue
orchestrator autonomous loop, CVE enrichment, the SSL/XSS/SQLi web scanners,
and persistence, scoring, ROE, ATT&CK, password and OpenAPI parsing.
Maintenance
v0.2.0 — Security hardening
A hardening release focused on trustworthiness: two real security fixes, green CI, supply-chain scanning, and a large jump in test coverage. Full details in CHANGELOG.md.
🔒 Security fixes
- Guardrail bypass fixed —
secsuite audit remediateexecuted AI-generated and operator-entered shell commands without the guardrails safety analyzer (the destructive-command blocklist:rm -rf,dd, fork bombs…). Every command now passes throughguardrails.validate_script()before execution. - API auth fixed — the
X-API-Keygate raisedHTTPExceptioninside middleware and surfaced as an uncaught 500 instead of 401. It now returns a proper 401, and usessecrets.compare_digestto prevent timing attacks.
➕ Added
- Supply-chain security tooling:
SECURITY.md, Dependabot, CodeQL, and a Security workflow (pip-audit+bandit). - 54 new tests. Coverage rose from 10.8% → 19.3% (guardrails 0→95.8%, compliance 0→98.8%, password 0→100%, SIEM 0→94.8%, API server 0→90.9%).
🔧 Changed
- CI lint is green (984 → 0
rufffindings);mypyruns again. - Exception chaining, dead-code removal, and timezone-aware datetimes.
Full Changelog: v0.1.0...v0.2.0