Abstract

Industrial Control Systems (ICS) are increasingly vulnerable to cyberattacks, making robust anomaly detection crucial for maintaining operational integrity and safety. This paper presents a novel hybrid deep learning framework designed to enhance anomaly detection capabilities in ICS environments. The framework combines the strengths of Convolutional Neural Networks (CNNs) for feature extraction from raw sensor data and Recurrent Neural Networks (RNNs), specifically Long Short-Term Memory (LSTM) networks, for capturing temporal dependencies within system behavior. By integrating these two architectures, the proposed model effectively learns complex patterns and detects subtle deviations indicative of anomalies. The framework is evaluated using a benchmark ICS dataset, demonstrating superior performance compared to traditional machine learning methods and single deep learning models. The results highlight the potential of the hybrid approach for improving the security and reliability of critical infrastructure.