PREDICTIVE THREAT MODELING IN INDUSTRIAL IOT (IIOT) NETWORKS USING MACHINE LEARNING TECHNIQUES IN CLOUD ENVIRONMENTS

Abstract

Cloud-enabled Industrial Internet of Things (IIoT) networks are being used more and more. These networks have changed how automation, tracking, and control are done in factories, but they have also made security much harder because they create so much different and changing data. There is an absolute need for predictive threat modeling in IIoT cloud environments because traditional signature-based solutions are notoriously bad at detecting new and unknown assaults.  In order to forecast potential threats in IIoT network traffic, this research suggests a machine learning-based strategy, using the CICIDS 2017 dataset as a baseline. After extensive preprocessing operations such as data cleaning, normalization, feature selection, and data balancing through SMOTE, a Convolutional Neural Network (CNN) was trained to automatically draw and understand complicated spatial-temporal patterns from multidimensional traffic data. Accuracy (ACC) was 99.23%, precision (PRE) was 98.32%, recall (REC) was 99.15%, and F1-score (F1) was 98.35%; this model outperformed its competitors, which included Logistic Regression (84.1%), LSTM (93.78%), and MLP (97.7%). proving that it can separate legitimate traffic from malicious ones. Findings show that predictive threat modeling based on deep learning is a good way to make IIoT networks in the cloud more secure and reliable