A Hybrid Simulated Annealing-XGBoost Framework with Feature Selection for Enhanced Network Intrusion Detection
Min Qi Zhou, Lichun Feng, Weijie Chen, Bingnan Li, Chuanxing Lin, Liqing Wang · 2025
In cybersecurity, evolving cyber-attack techniques demand higher accuracy and efficiency from intrusion detection systems. To effectively tackle this challenge, we propose a hybrid framework named SAXGB - FS, which integrates the global search ability of Simulated Annealing, the predictive power of XGBoost, and a specialized feature selection strategy to enhance the performance of network intrusion detection. Firstly, a feature selection strategy that combines the chi-square test and mutual information analysis is adopted to select the most discriminative features, thereby reducing the complexity of model training. Secondly, leveraging the global optimization property of the Simulated Annealing algorithm, we optimize the key parameters of the XGBoost model. Finally, the optimized hybrid framework is applied to the dataset on Kaggle for experimental validation. The results indicate that the proposed hybrid framework achieves an accuracy of $99.64 \%$. Compared with other algorithms, our approach demonstrates significant advantages in various performance metrics. This provides an efficient and precise technical solution for cybersecurity protection.