ICSentinel Modbus Autoencoder
PyTorch autoencoder trained on benign Modbus TCP traffic for OT/ICS anomaly detection.
Model Details
| Property | Value |
|---|---|
| Architecture | Autoencoder (10 → 32 → 16 → 8 → 16 → 32 → 10) |
| Features | 10-dim Modbus frame vector |
| Training frames | 180 benign (normal_operations + normal_polling) |
| Threshold | 99th percentile reconstruction error |
| Version | 1.0.0 |
Intended Use
- OT/ICS Modbus TCP anomaly scoring
- Demo and evaluation on synthetic traffic only
- Not for production OT IDS replacement without retraining on site-specific baselines
Limitations
- CPU inference only in demo
- Synthetic training data — retrain on local benign traffic for production
- Rule-based attack classification supplements autoencoder scores
Built by Aria AI.
- Downloads last month
- 23
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support