Integrated Power System Cybersecurity Enhancement Through Gegenbauer Graph Neural Networks Optimized With Multiagent Cubature Kalman Optimizer for Reliable Attack Prevention

ABSTRACT The increasing complexity and interconnectivity of modern smart grids, particularly through SCADA systems and IoT networks, have heightened vulnerabilities to cyberattacks such as false data injection, relay setting changes, and tripping command attacks. These threats can lead to delayed detection, infrastructure damage, financial losses, and even fatalities. To overcome these issues, Integrated Power System Cybersecurity Enhancement through Gegenbauer Graph Neural Networks (GGNN) Optimized with Multiagent Cubature Kalman Optimizer (MACKO) for Reliable Attack Prevention (GGNN‐CSPS‐MACKO) is proposed. Initially, the input data is gathered from the power system attack detection dataset. The gathered data is preprocessed using the Regularized bias‐aware Ensemble Kalman Filter (RB‐AEKF) to clean the data frames. Then, the cleaned data is fed into GGNN for classifying the cyber security threats in power systems as Natural events and Attack events. Afterward, the MACKO is applied to optimize the weight parameters of the GGNN classifier to enhance accuracy. The proposed IPCDT‐AMCGNN‐VC method is implemented and its efficiency was evaluated using several performance metrics like accuracy, precision, F1‐score, recall, ROC, Loss, and True Negative rate. The proposed GGNN‐CSPS‐MACKO approach has 25.10%, 20.45%, and 17.20% higher accuracy when analyzed to the existing methods CSPS‐MH‐DIA, DRTMI‐IoT‐CSI, and MADDL‐DCA‐DR.