A Novel Approach for Intrusion Detection in Cyber-Physical Systems Using Decision Tree and Feature Dimensionality Reduction

Cyber-Physical Systems (CPS) have become increasingly essential in critical infrastructure, yet they remain highly vulnerable to sophisticated cyber threats. Traditional intrusion detection techniques—particularly signature-based and anomaly-based systems—often fail to detect emerging attacks and tend to suffer from high false positive rates. In this study, we propose a novel intrusion detection method that integrates Principal Component Analysis (PCA) for dimensionality reduction with a decision tree classifier. This hybrid model enhances detection accuracy while reducing computational overhead. Using the KDD-99 dataset, the proposed approach achieved a detection accuracy of 95.06%, outperforming conventional methods such as ID3 and UCSm. The method also demonstrated superior performance in terms of reduced processing time and lower resource consumption. These results suggest that the proposed system is highly effective for resource-constrained CPS environments where fast and reliable threat detection is crucial.