Characteristic Canonical Analysis-Based Attack Detection of Industrial Control Systems in the Geological Drilling Process

Modern industrial control systems (ICSs), which consist of sensor nodes, actuators, and buses, contribute significantly to the enhancement of production efficiency. Massive node arrangements, security vulnerabilities, and complex operating status characterize ICSs, which lead to a threat to the industrial processes’ stability. In this work, a condition-monitoring method for ICSs based on canonical variate analysis with probabilistic principal component analysis is proposed. This method considers the essential information of the operating data. Firstly, the one-way analysis of variance method is utilized to select the major variables that affect the operating performance. Then, a concurrent monitoring model based on probabilistic principal component analysis is established on both the serially correlated canonical subspace and its residual subspace, which is divided by canonical variate analysis. After that, monitoring statistics and control limits are constructed. Finally, the effectiveness and superiority of the proposed method are validated through comparisons with actual drilling operations. The method has better sensitivity than traditional monitoring methods. The experimental result reveals that the proposed method can effectively monitor the operating performance in a drilling process with its highest accuracy of 92.31% and a minimum monitoring delay of 11 s. The proposed method achieves much better effectiveness through real-world process scenarios due to its distributed structural division and the characteristic canonical analysis conducted in this paper.