网络安全与数据治理2025，Vol.44Issue(5)：1-9,16,10.DOI:10.19358/j.issn.2097-1788.2025.05.001
基于可解释LightGBM的电动汽车充电站入侵检测方法
Intrusion detection method for electric vehicle charging station based on interpretable lightweight gradient boosting machine
摘要
Abstract
Against the backdrop of increasingly severe cybersecurity challenges in Electric Vehicle Charging Stations(EVCS),traditional intrusion detection methods exhibit multiple limitations,while machine learning and deep learning approaches,despite their effectiveness,suffer from"black-box"issues.This paper proposes an interpretable Lightweight Gradient Boosting Machine(LightGBM)-based intrusion detection framework for EVCS.The framework employs SHAP for feature selection and utilizes a Simulated Annealing Arithmetic Optimization Algorithm(SAOA)to optimize LightGBM hyperparameters,while integrating multi-ple Explainable Artificial Intelligence(XAI)techniques including SHAP,LOCO,CEM,PFI,LIME,and ALE.Experimental results on the CICEVSE2024 and Edge-IIoTset datasets demonstrate that the model achieves detection accuracies of 97.53％and 88.89％,with F1-scores of 98.01％and 88.98％respectively,while maintaining strong interpretability to provide clear decision-making basis for security operations.This research offers an efficient and interpretable solution for enhancing EVCS cybersecurity,with significant theoretical and practical implications.关键词
电动汽车充电站/入侵检测/轻量级梯度提升机/可解释人工智能Key words
electric vehicle charging stations/intrusion detection/LightGBM/XAI分类
电子信息工程引用本文复制引用
姚沁怡,龙莆均,陈世伦..基于可解释LightGBM的电动汽车充电站入侵检测方法[J].网络安全与数据治理,2025,44(5):1-9,16,10.基金项目
重庆市教委科学技术研究项目(KJQN202101536) （KJQN202101536）
重庆科技大学硕士研究生创新计划项目(YKJCX2321110) （YKJCX2321110）
