DEIDS: a novel intrusion detection system for industrial control systems

Abstract Owing to the development of industrial production, hidden danger in control systems (ICSs) has considerably increased, causing challenges traditional safety defense methods. The combination machine-learning or deep-learning algorithms and intrusion detection (IDSs) become mainstream method for solving this problem. However, these methods depend on a massive amount high-quality attack traffic data, which cannot be obtained easily owing independence unique characteristics ICSs. In study, we apply reconstructed convolutional neural network data expansion algorithm named CenterBorderline_SMOTE ( CB_SMOTE ) an IDS propose system DEIDS ). is end-to-end model that learns representative features from raw classifies them unified framework. Moreover, adopt classification activation map structure, can deeply mine potential enhance effectiveness features. While enhancing quality, introduce designed into expand solve problem insufficient system. Our comprehensive experiments different open datasets indicate achieves excellent performance (97 $$\%$$ % accuracy) outperforms state-of-the-art experimental results also show our high efficiency accuracy processing ICSs datasets.