A Machine learning based intrusion detection approach for industrial networks

Industry 4.0 and Industrial Internet of Things (IIoT) are current trends in the industrial automation world. They require connections of factory networks to the internet. This trend increases the vulnerability of factory networks to attacks. Here, we present an approach that monitors the activities of factory network traffic based on two linear feature extraction algorithms, i.e. LDA and PCA. A Machine-Learning-based approach is used to analyze the records of network connections from the UNSW-NB15 database and to detect and report anomalies such as malicious attacks. The experimental results show the feasibility of the provided method in accuracy, detection rate, and false alarm rate.