Topological Data Analysis-Based Replay Attack Detection for Water Networks

Cyber attacks have become a critical problem over the years. A specific cyber-attack called replay attack is known to be hard to detect and it can cause a serious damage on cyber-physical systems (CPSs). In this paper, a replay attack detection that uses Topological Data Analysis (TDA) is considered for sensor signals of water distribution networks. TDA is an effective approach to capture the periodicity information of the sensor measurements, which is a strong motivation for replay attack detection. Several cases of replay attacks are generated. TDA is modified to deal with time series - the measurements. The main idea is to learn the topological features of the nominal measurements and compare them with the disrupted measurements. To understand the periodicity information of the data, the Betti curves are used to keep track on the topological features.