An Anomaly Detection and Scenario Classification Scheme Based on Fuzzy C-means Clustering

In this paper, problems of anomaly detection and scenario classification in industrial control systems(ICSs) are investigated. Anomalies caused by attacks can have impacts on product quality, production stability and device security in ICSs to varying degrees, where different countermeasures are needed. To distinguish anomalies with different damage, the rationale of ICSs is analyzed in detail and four security scenarios are defined with typical characteristics, taking Tennessee Eastman process as an example. A set of attributes are extracted from typical data of scenarios and fuzzy c-means algorithm is adopted to classify sample cases into these security scenarios. At last, an anomaly detection and scenario classification scheme is proposed with a data-driven security scenario model. Experiments are provided to verify the validity and generality of the proposed method.