Multi-Level Decision and Adaptive Defense Against DoS Attacks in Smart Grids

Smart grids face substantial challenges in terms of cybersecurity, especially from denial-of-service (DoS) attacks, which represent a critical threat to system stability. To address this pressing issue, this paper proposes a multi-level decision detection and adaptive defense approach specifically designed to mitigate DoS attacks. Firstly, this method integrates Lasso regression, the XGBoost classifier, and random forest algorithms to extract key network traffic features, thereby improving the accuracy of attack detection. Secondly, it uses support vector machines (SVM) and decision trees for preliminary classification, with XGBoost serving as the final classifier to build a multi-level decision model. In order to comprehensively con-sider multiple performance metrics, the hyperparameters of the model are adjusted through multi-objective Bayesian optimization, enabling real-time monitoring and precise identification of net-work traffic patterns, effectively detecting DoS attacks. Upon dete-ction, the system promptly deploys adaptive defense strategies to mitigate the impact. Experimental results show that this method significantly enhances the detection accuracy and response effi-ciency of DoS attacks, effectively ensuring the stability and sec-urity of smart grids.