Machine Learning Techniques for Anomaly Detection in the Hydra Testbed: A Data-Driven Defense Strategy

As cyber-attacks targeting Critical Infrastructures become increasingly sophisticated, successfully identifying breaches has become more challenging. Failure to detect intrusions can undermine the confidence in security services, compromising data confidentiality, integrity, and availability. This paper introduces the HydraCPS dataset, a novel resource designed to detect cyber-physical attacks on water distribution systems. Generated through simulations on the Hydra testbed, which emulates a real-world water distribution system, the dataset includes ground truth labels essential for supervised training. Intrusion detection is framed as a classification problem, employing various Artificial Intelligence techniques to distinguish between nominal and attack conditions. The comprehensively labeled HydraCPS dataset provides a robust foundation for IDS evaluation. The Hydra testbed’s detailed simulation of a water distribution system makes it an effective tool for identifying attacks on critical infrastructure. This enhances the relevance and applicability of the HydraCPS dataset in developing and evaluating IDS models aimed at protecting vital systems. Our goal is to facilitate the design of efficient and effective intrusion detection systems by leveraging new IDS datasets and utilizing the detailed and realistic attack simulations provided by HydraCPS for improved model training and evaluation.