Skip to content

Repository files navigation

GuardIAn — Detection & Response Platform

Hackathon Togo IT Days 2025 — EPL Team — v2 Refonte

GuardIAn is an open-source EDR-style (Endpoint Detection & Response) platform focused on ransomware and other commodity malware. It is composed of three independent components designed to scale from a single workstation to a small SOC:

Component Role Tech
Agent Runs on each protected Windows endpoint. Watches files, processes and registry. Computes lightweight heuristics (entropy, signatures, YARA) and forwards events to the backend. Python 3.12 + watchdog + psutil + (optional) PyInstaller
Backend Central API. Receives events, runs the unified ML detector + rule engine (YARA/Sigma), persists threats, exposes REST + WebSocket. FastAPI + SQLAlchemy + PostgreSQL + Redis (RQ) + scikit-learn
Frontend SOC analyst dashboard: live status, threats, scans, eradication plans, intel. React 18 + Vite + TypeScript + Tailwind + Recharts
flowchart LR
  subgraph EP["Windows Endpoint"]
    A[guardian-agent]
  end
  subgraph SRV["Linux Server (Docker)"]
    NG[nginx + TLS] --> API[FastAPI]
    NG --> FE[React UI]
    API --> DB[(PostgreSQL)]
    API --> RD[(Redis queue)]
    RD --> WK[ML Worker]
    WK --> DB
  end
  A -->|HTTPS + JWT| NG
  ANALYST[SOC analyst] -->|HTTPS| NG
Loading

Quick start (development, Linux/macOS/Windows + Docker Desktop)

git clone https://github.com/Patrickleondev/Hackathon_au_Togo_IT_Days_2025_EPL_Team.git
cd Hackathon_au_Togo_IT_Days_2025_EPL_Team
cp infra/.env.example infra/.env
# Edit infra/.env and set SECRET_KEY, BOOTSTRAP_ADMIN_PASSWORD, POSTGRES_PASSWORD
docker compose -f infra/docker-compose.yml up --build

Then open:

Default analyst credentials come from BOOTSTRAP_ADMIN_EMAIL / BOOTSTRAP_ADMIN_PASSWORD in infra/.env.

For full prerequisites, local/VPS choices, troubleshooting and production hardening, start with docs/PREREQUISITES.md, docs/04-INSTALL.md and docs/08-DEPLOYMENT.md.


Documentation

All detailed docs live under docs/:

File Topic
01-VISION.md Problem statement, threat model, scope, non-goals
02-ARCHITECTURE.md Components, data flow, network topology, sequence diagrams
03-ALGORITHMS.md Detector internals (features, scoring, rule engine, eradication)
04-INSTALL.md Local install, development modes, Docker Compose, troubleshooting
05-AGENT-WINDOWS.md Building, deploying and operating the Windows agent
06-API.md REST + WebSocket reference (companion to /docs OpenAPI)
07-OPS.md Logs, metrics, backups, rule updates, troubleshooting
08-DEPLOYMENT.md VPS/on-prem production, TLS, secrets, hardening, backups
09-CONTRIBUTING.md Branching, code style, tests, release process

Repository layout

.
├── agent/          # Windows endpoint agent (Python)
├── backend/        # FastAPI server + ML worker
├── frontend/       # Vite + React SOC dashboard
├── infra/          # docker-compose, nginx, env templates
├── docs/           # All architecture / install / ops docs
├── scripts/        # Training, samples, utilities
├── legacy/         # Original Hackathon 2025 codebase (frozen, not built)
├── LICENSE
└── README.md

The directory legacy/ keeps the original hackathon code as a historical reference. It is not built, not deployed, and not exposed.


Security posture

  • Local-only by default — no telemetry leaves the operator's network.
  • mTLS-ready between agent and backend (see docs/08-DEPLOYMENT.md).
  • No payload ever runs server-side — analysis is performed on hashes, metadata, and sandboxed feature extraction only.
  • Deny-by-default quarantine: dry-run is always the default for eradication plans (POST /api/eradications requires dry_run=false explicitly to take action).

License

MIT — see LICENSE.


🇹🇬 Credits

Originally built during the Togo IT Days 2025 hackathon by the EPL Team. v2 refonte focuses on producing a coherent, deployable, documented, production-grade reference implementation.

CERT-TG hotline (incident response in Togo): (+228) 70 54 93 25

About

développement de solutions de protection contre les menaces de ransomware à l'aide de l'Intelligence Artificielle

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages