Defensive SOC investigation + detection engineering workbench.
Normalize telemetry, evaluate YAML detections, correlate process activity, extract IOCs, and review MITRE ATT&CK coverage.
Release:
v0.2.0 — Telemetry Engine
SentinelForge is a portfolio/open-source defensive security project. It is not a SIEM replacement and does not claim full Sigma compatibility.
SOC portfolios often say “I know Sysmon, Wazuh, MITRE ATT&CK and incident investigation.” SentinelForge turns those claims into an inspectable project: parsers, a normalized event model, detection rules, correlation logic, tests, CI, a CLI, an API, and an analyst-facing interface.
- Windows
.evtxingestion throughpython-evtx - Wazuh/Sysmon-friendly normalization into a common event schema
- External YAML detection rules instead of Python-hardcoded matches
- 12 curated defensive detections with current MITRE ATT&CK mappings
- Severity and confidence-aware risk scoring
- IOC extraction for URLs, domains, email, IPv4, MD5, SHA-1, SHA-256
- PID/PPID process graph correlation
- ATT&CK technique hit summary
/api/rulesdetection inventory- CLI analysis + JSON export
- Hardened container baseline and response security headers
- Multi-version CI, Ruff linting, tests, Dependabot
- Redesigned analyst workspace with drag/drop, filters and export
Run the service, click Load demo, then Run analysis. The bundled synthetic dataset models a suspicious document → PowerShell → download → child process → scheduled task/defense-tampering chain.
WINWORD.EXE (PID 4080)
└── powershell.exe (PID 5116)
└── agent.exe (PID 6240)
├── schtasks.exe (PID 6352)
└── powershell.exe (PID 6404)
The dataset is synthetic and uses documentation-only domains/IP ranges.
git clone https://github.com/ShalvaLekishvili/SentinelForge.git
cd SentinelForge
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements-dev.txt
make devOpen http://127.0.0.1:8000.
docker compose up --buildpython -m backend.cli sample-data/demo-incident.json --output investigation.jsonTelemetry
│
├── EVTX parser
├── JSON / Wazuh parser
├── CSV parser
└── line-oriented parser
│
▼
NormalizedEvent
│
├── YAML detection engine ──► findings + MITRE mappings
├── IOC extractor ──► indicators
├── PID/PPID correlation ──► process graph
└── scoring ──► risk summary
│
▼
FastAPI / Analyst UI / CLI
See docs/ARCHITECTURE.md for design details.
Rules live in rules/ and are intentionally Sigma-inspired: metadata, log source context, detection clauses, condition logic, and ATT&CK mappings are separate from Python code.
| Rule | Detection | Severity | ATT&CK |
|---|---|---|---|
| SF-EXEC-001 | Encoded PowerShell | High | T1059.001 |
| SF-EXEC-002 | Office → PowerShell | High | T1204.002, T1059.001 |
| SF-PERS-001 | Registry Run/RunOnce | Medium | T1547.001 |
| SF-PERS-002 | Scheduled Task creation | Medium | T1053.005 |
| SF-CRED-001 | Potential LSASS memory dump | Critical | T1003.001 |
| SF-DEF-001 | Defender real-time monitoring tampering | High | T1562.001 |
| SF-TRANSFER-001 | PowerShell remote file transfer | High | T1105, T1059.001 |
| SF-LOLBIN-001 | Rundll32 suspicious proxy execution | High | T1218.011 |
| SF-LOLBIN-002 | Regsvr32 remote scriptlet pattern | High | T1218.010 |
| SF-LOLBIN-003 | Mshta remote/script execution | High | T1218.005 |
| SF-PERS-003 | Service creation via sc.exe |
Medium | T1543.003 |
| SF-NET-001 | PowerShell with destination IP | Medium | T1059.001 |
Read docs/RULE_AUTHORING.md before adding detections.
| Endpoint | Method | Purpose |
|---|---|---|
/api/health |
GET | Service version + active rule count |
/api/rules |
GET | Detection rule inventory |
/api/analyze |
POST | Analyze EVTX/JSON/CSV/TXT/LOG telemetry |
/docs |
GET | OpenAPI / Swagger UI |
backend/
parsers/ EVTX + structured/text parsers and normalization
services/ detection, IOC, scoring, correlation, analysis
cli.py terminal analysis entry point
frontend/ dependency-free analyst UI
rules/ external YAML detection-as-code library
sample-data/ synthetic investigation datasets
tests/ parser, engine, API, correlation, scoring tests
docs/ architecture, rules, roadmap, demo notes
assets/ README visual assets
.github/ CI, Dependabot, issue templates
scripts/ GitHub Project bootstrap helper
- Explainable detections — every hit carries rule, evidence, severity, confidence and ATT&CK context.
- Small core — avoid hiding core investigation logic behind a heavy framework.
- Portable telemetry — normalize multiple source shapes before detection.
- Detection-as-code — rules are reviewable data files with tests.
- Synthetic-by-default — public demos must not expose real incident artifacts.
- Defensive scope — analysis and detection capabilities, not exploitation automation.
make install
make lint
make testGitHub Actions runs lint + tests across supported Python versions. Pull requests that add detection rules should include a positive test/sample and document expected false positives.
- v0.3 — richer Sysmon Event ID semantics, field aliases, investigation search, rule tests
- v0.5 — persisted cases, analyst notes, evidence tagging, timeline filters
- v0.7 — Sigma import/validation experiment, Wazuh adapter, HTML report export
- v1.0 — stable schema, release docs, larger synthetic investigation pack
Full plan: docs/ROADMAP.md.
Do not upload sensitive production logs to public demos. SentinelForge does not need outbound internet access for its core analysis path. See SECURITY.md.
MIT © 2026 Shalva Lekishvili