Abstract
Rapid technological advancements have led to the widespread deployment of wireless sensor networks (WSNs) in industrial environments, making cybersecurity a critical concern in cloud computing. This paper presents a predictive framework for cloud-based intrusion detection and prevention for WSNs. It integrates machine learning models- Multilayer Perceptron (MLP), Decision Tree, and Autoencoder-to precisely classify and mitigate various impacts of cyber intrusions on a cluster of wireless sensors. An intelligent prioritization and prevention system is also proposed, categorizing attacks-blackhole, grayhole, flooding, and scheduling-based on their impact on industrial processes. Experimental results indicate robust detection capabilities, with the Decision Tree achieving 99.48% accuracy, slightly outperforming MLP at 99.37%. The Autoencoder demonstrated superior binary classification, distinguishing between normal and anomalous instances with high precision and recall rates. This framework leverages the WSN-DS dataset to simulate and validate its efficiency in mitigating real-time threats. Future work will focus on refining the prioritization model and integrating advanced machine learning techniques for enhanced adaptability and resilience.