SWat AI Project

This project proposes the development of an advanced anomaly detection system for industrial equipment. The system will use self-supervised learning, specifically a contrastive learning approach, to analyze multivariate time-series data from IoT sensors.

Instead of requiring historical data of failures, the model will learn the signature of normal operational behavior from unlabeled data. By identifying any deviations from this learned norm, the system can provide early warnings for potential equipment failures or cyber-physical attacks, enabling proactive maintenance and enhancing security.

For implementation and testing, this project can utilize the SWaT (Secure Water Treatment) dataset, which contains extensive real-world sensor data with labeled attack scenarios, making it ideal for validating the model's effectiveness. This project is highly relevant to the Industry 4.0 landscape and offers significant potential for a strong research paper and a practical application in predictive