Abstract
The fourth industrial revolution's transformation utilizes a Cyber-Physical System (CPS) to secure Supply Chain 4.0. It integrates manufacturing information with internet communication technology to create a smart CPS that tracks products from manufacture to customer delivery using the Internet of Things (IoT). This research uses a Machine Learning (ML) approach for network anomaly detection and constructing data-driven models to detect DDoS attacks on Industry 4.0 CPSs. Limitations of existing techniques, such as artificial data and small datasets, are addressed by capturing network traffic data from a real-world semiconductor production factory. 45 bidirectional network flow features are extracted, and labeled datasets are constructed for training and testing ML models. The proposed PCA-BSO algorithm is employed to select the most relevant features based on their eigenvalues, as the feature with the highest eigenvalues may not always improve classification accuracy. Supervised ML algorithms are evaluated through simulations to assess their performance.