AI-powered, multi-modal cybersecurity threat detection with MITRE ATT&CK integration, adversarial robustness testing, and enterprise SIEM support.
| Feature | Description | Status |
|---|---|---|
| 📧 Phishing Detection | DistilBERT-based email analysis | ✅ |
| 🔐 LLM Security | Prompt injection detection (Zelis AI compliant) | ✅ |
| 📊 Log Anomaly | Isolation Forest for anomaly detection | ✅ |
| 🌐 Network Security | Real-time packet capture | ✅ |
| 🎯 MITRE ATT&CK | Complete v14.0 mapping | ✅ |
| 🛡️ Adversarial Robustness | FGSM/PGD attack simulation | ✅ |
| 📤 SIEM Integration | Microsoft Sentinel + Slack/Email | ✅ |
| 🔧 CI/CD Security | Bandit + Safety + Trivy | ✅ |
DeepGuard AI/ ├── 📧 Phishing Detector (DistilBERT) ├── 📊 Log Anomaly Detector (Isolation Forest) ├── 🌐 Network Anomaly Detector (Isolation Forest) ├── 🔌 FastAPI Backend └── 💻 Streamlit Frontend
text
- Clone Repository
git clone <repository-url>
cd DeepGuard-AI
Install Dependencies
bash
pip install -r requirements.txt
Train Models
bash
python train_models.py
Start API Server
bash
python run_api.py
Start Frontend (New Terminal)
bash
streamlit run frontend/app.py
Access System
API Docs: http://localhost:8000/docs
Dashboard: http://localhost:8501
📊 API Endpoints
POST /analyze/phishing - Analyze emails for phishing
POST /analyze/logs - Detect anomalies in system logs
POST /analyze/network - Monitor network traffic
POST /analyze/comprehensive - Complete security scan
GET /system/status - System health check
🧠 ML Models
Phishing Detection: Fine-tuned DistilBERT model
Log Analysis: Isolation Forest for anomaly detection
Network Analysis: Isolation Forest for traffic patterns
🎯 Usage Examples
Phishing Detection
python
import requests
response = requests.post("http://localhost:8000/analyze/phishing",
json={"emails": ["You won $1000! Click here..."]}
)
print(response.json())
Comprehensive Scan
python
payload = {
"emails": [...],
"logs": [...],
"network_traffic": [...]
}
response = requests.post("http://localhost:8000/analyze/comprehensive", json=payload)
📈 Performance
Phishing Detection Accuracy: ~90%
Log Anomaly Detection: ~85%
Network Threat Detection: ~88%
Response Time: < 2 seconds
🔧 Development
Backend: FastAPI + Python
Frontend: Streamlit
ML: PyTorch, Scikit-learn, Transformers
Data: Pandas, NumPy