Cyber-Physical System Security Using AI-Based Intrusion Detection And Predictive Threat Intelligence Models

Abstract

A new concept of Cyber-Physical Systems (CPS) has become an essential element of the current industrial automation, intelligent healthcare, intelligent transportation, and intelligent grid systems. Nevertheless, with the growing connection of physical devices to communication networks, vulnerability to advanced cyber threats, such as distributed denial of service attacks, malware injection and data manipulation, and unauthorized access have greatly increased. Traditional intrusion detection tools have repeatedly been shown to offer few capabilities in monitoring dynamic attacks and zero-day attacks in the dynamic CPS environment. In this paper, an AIbased intrusion detection and predictive threat intelligence taskforce is suggested to boost CPS cybersecurity. The proposed framework incorporates combination of a hybrid Convolutional neural network- Long short term memory ( CNN- LSTM) model and predictive threat intelligence to enhance accuracy of attack detection and proactive threat prediction. The CNN component learns spatial traffic features and LSTM network learns temporal attack behaviors based on the CPS network traffic data. The CICIDS2017 dataset with various actual attack types was used to evaluate the experiment. The proposed framework obtained an intrusion detection accuracy of 97.2, precision of 96.8, recall of 96.1, F1-score of 96.4 and ROC-AUC score of 0.982. Comparison revealed high performance in comparison to traditional machine learning and stand-alone deep learning models. The proposed framework was statistically validated using 10-fold cross-validation to verify the strength and the generalization of the proposed framework. The designed predictive threat intelligence module also enhanced the proactive prediction of attacks and cybersecurity responsiveness on CPS infrastructures. The proposed framework offers an extensible and smart security implementation to next generation Cyber-Physical Systems.