This project implements an AI-driven SIEM system that collects and monitors SSH honeypot activity (via Cowrie), detects anomalies using machine learning, visualizes data through a real-time dashboard, and sends alert notifications via email.
- 🔐 SSH honeypot via Dockerized Cowrie
- 📦 Logs shipped to Elasticsearch using Filebeat
- 📊 Real-time anomaly detection with Isolation Forest
- 📈 Streamlit dashboard with live plots and filters
- 📬 Email alerts using Brevo API (SMTP)
- 🧠 Automatically launches dashboard on detection
- Cowrie – SSH honeypot
- Docker + ELK Stack – Log ingestion and search
- Python – Data processing and ML
- Scikit-learn – Anomaly detection
- Plotly – Data visualization
- Streamlit – Interactive dashboard
- Brevo – Email alerting
cowrie/
├── docker-compose.yml # Deploys Cowrie + Elasticsearch + Kibana
├── filebeat-config.yml # Configures log shipping
├── cowrie-logs/ # Cowrie log output
├── session-logs/ # Captured attacker sessions
ai-engine/
├── detect_anomalies.py # Runs ML detection, saves anomalies
├── app.py # Streamlit dashboard UI
├── anomalies.csv # Anomaly output
├── requirements.txt # Python dependencies
alerting/
├── alert_manager.py # Sends Brevo email alerts
cd cowrie
docker-compose up -d- Cowrie logs are saved locally and shipped to Elasticsearch using Filebeat.
- Confirm logs are visible in Kibana or via Elasticsearch queries.
pip install -r ai-engine/requirements.txtpython ai-engine/detect_anomalies.py- Performs log query, processes features, and applies anomaly detection
- If anomalies are detected:
- Saves to
anomalies.csv - Triggers Streamlit dashboard (if not already open)
- Sends email alerts via Brevo
- Saves to
streamlit run ai-engine/app.py- Configure your SMTP/Brevo credentials inside
alerting/alert_manager.py - Alerts are triggered automatically on detection of anomalies
- Model: Isolation Forest
- Features Used:
- Time of day
- Command frequency
- IP frequency
- Session duration
- Login attempts
- Top IPs and session stats
- Command pattern visualizations
- Auto-refreshing plots
Victor David Sarkibaka
GitHub | LinkedIn | victor.baka16@gmail.com