面向无人机网络攻击检测的两级特征选择方法OA

To address the dual challenges of high computational demands in machine learning and the limited computing resources of unmanned aerial vehicles(UAVs),as well as the drawback of traditional fixed-binning information gain feature selection methods which underestimate highly discriminative features,this paper proposes a UAV intrusion detection scheme based on a two-level feature selection approach.The scheme adopts a"Chi-square test preliminary screening-Heuristic Information Gain Feature Selection(HIS)fine selection"method,which a-daptively determines the optimal binning for each feature to accurately quantify its discriminative power.Meanwhile,the number of features is treated as a hyperparameter and jointly optimized with the XGBoost classifier.Experiments on the UAV-NIDD dataset demonstrate that the pro-posed method maintains high detection performance while reducing model detection time by approximately 97.5%.The results verify that this scheme effectively balances detection accuracy and computational cost,providing an efficient and real-time intrusion detection capability for re-source-constrained UAV platforms.