WH-XGBoosting: A Multi-Stage Intrusion Detection Framework for Securing Communication in Electric Vehicle Smart Grid Networks

Abstract

Electric vehicles (EVs) are mostly linked with the smart grids that cause diverse cyberattacks such as denial of services (DoS), data manipulations and network intrusions, which affect the grid ecosystem's reliability, efficiency and security. The multi-stage intrusion detection framework is created to explore the various resources, power consumption metrics, and network traffic to identify and mitigate cyberattacks. The adoption of EVs in grid systems creates dynamic security issues and complexity while exchanging information. The research difficulties are addressed by developing the whale-optimised XGBoosting machine learning (WH-XGBoosting), which can identify and mitigate the threats by attaining scalability and low latency. The framework uses diverse features and segmentation procedures to reduce redundancy and overfitting issues. In addition, the whale optimisation process selects optimised values and hyperparameters that improve the detection rate. Then, a boosting algorithm is applied to classify the incoming data, with a minimum false positive rate and maximum detection rate. The framework uses the whale optimisation process to select the optimized features and classifier hyperparameter updating process that enhance the overall intrusion detection accuracy. The discussed system collects the input from CICEVSE2024 and processes it using high-level feature analysis, which helps predict the intruder with a maximum recognition rate (99.12%) compared to existing methods. The system ensures robust, reliable, and scalable solutions for various cyber threats in grid ecosystems.