INTELLIGENT RANSOMWARE DETECTION IN INDUSTRIAL CONTROL NETWORKS USING EFFICIENT MACHINE LEARNING MODELS

Abstract

The development of advanced cyberattacks, including targeted ransomware, requires more stringent security protocols,
and this gave way to the introduction of zero-trust (ZT) deployment as a solution to these issues. To overcome this difficulty, this
research proposes an effective ransomware detection model that leverages ransomware data via a deep learning-based Gated
Recurrent Unit (GRU). The achieved GRU model performed better with an accuracy (ACC) of 98.53%. Compared to traditional
models such as CNNs, SVMs, and Logistic regression, the dominance of GRUs in reducing false predictions and in capturing
time-related patterns was evident in the comparative analysis. Stable convergence was confirmed by the training and validation
curves, and the confusion matrix showed only a few misclassifications. Optimized feature selection also increases detection ability
by focusing on the most pertinent features. These findings confirm the GRU model's strength, generalization ability, and
applicability in industrial control systems to reduce ransomware attacks. On the whole, the suggested solution provides a flexible,
stable scheme for ransomware detection that can be deployed in industrial control systems and critical infrastructure settings where
timely, accurate threat detection is the priority.