A Traffic Analysis and Node CategorizationAware Machine Learning-Integrated Framework for Cybersecurity Intrusion Detection and Prevention of WSNs in Smart Grids

Abstract

Smart grids are transforming the generation, distribution, and consumption of power, marking a revolutionary step forward for contemporary energy systems. Communication in smart grid environments is majorly performed through Wireless Sensor Networks (WSNs). The WSNs enable real-time monitoring and management inside smart grids. However, the integration of digital technologies and automation in smart grids introduces cybersecurity challenges, including unauthorized access, data breaches, and denial of service attacks. To address these difficulties and maintain the reliability of smart grid infrastructure, this study proposes a comprehensive architecture for strengthening cybersecurity within WSNs operating in smart grid environments. By integrating traffic analysis, node categorization, and machine learning algorithms, the framework intends to effectively detect and prevent cyber threats. Extensive evaluation reveals that traffic analysis using the Random Forest model successfully predicts traffic load within WSNs, achieving a mean squared error (MSE) of 2.772350, a root mean squared error (RMSE) of 1.665038, a mean absolute error (MAE) of 1.099080, and a coefficient of determination (R2 ) of 0.717982. In intrusion detection, the Random Forest model outperforms Decision Trees and Logistic Regression, with higher precision (0.99), recall (0.99), and F1 scores (0.98) across various attack types. Specifically, Random Forest achieves perfect recall (1.00) in identifying Flooding attacks, underscoring its capability to detect all instances of such intrusions. Leveraging the insights gathered from the WSNBFSF dataset, this study gives significant findings into proactive cybersecurity tactics, stressing the necessity of securing key infrastructure for the reliable and secure distribution of power to consumers.