Author : Ruchi Patel

Industrial Control Systems (ICS) prove more susceptible to cyber threats which makes it necessary to create effective threat detection systems. The improvements in cybersecurity fields do not deliver sufficient scalability with real-time threat detection functionality. The paper designs an AI-based framework for ICS cybersecurity defense which applies deep learning to automate threat discovery along with risk reduction procedures. This research explores machine learning While deep learning (DL) models detect cyber threats using CICIDS-2017 dataset information. The testing phase included a CNN primary classification model against KNN traditional models and Naïve Bayes (NB) traditional models together with Support Vector Machine (SVM) traditional models. The CNN model exhibited the best performance by reaching 99.58% accuracy and precision as well as recall and F1-score resulting in its well-documented superiority in detecting cyber threats. The model achieved verification of its performance by examining accuracy curves and loss diagrams together with confusion matrix results. Deep learning has proven its effectiveness in industrial control system security by delivering sustainable real-time invasion detection capabilities along with risk management solutions.

Keywords: Cybersecurity, Cyber Threats, Industrial Control Systems (ICS), Threat Detection, Risk Mitigation, Deep Learning (DL), CICIDS-2017.