【摘要】 入侵检测作为一种网络主动防御技术，能够有效阻止来自黑客的多种手段攻击。随着机器学习的发展，相关技术也开始应用到入侵检测中。本文采用sklearn库中preprocessing模块的函数对KDD CUP 99数据集进行预处理，基于朴素贝叶斯和逻辑回归算法，建立了网络入侵检测模型，并利用信息增益算法对入侵相关特征进行选择，然后进行训练与预测。实验结果表明，选择特征子集进行训练和预测能够保证预测准确率并大幅提高检测效率。研究成果可为高速铁路信号系统网络入侵检测模型的设计和建立提供参考。更多还原

【Abstract】 Intrusion detection, as an active defense mechanism in networking, effectively thwarts diverse forms of attacks by hackers. With the advancements in machine learning, related technologies are increasingly being employed in intrusion detection systems. This study utilized preprocessing functions from the sklearn library’s preprocessor module to preprocess the KDD CUP 99 dataset. Based on Naive Bayes and logistic regression algorithms, a network intrusion detection model was constructed, followed by feature selection using the information gain algorithm prior to training and prediction. Experimental results demonstrate that training and predicting with a subset of selected features ensures prediction accuracy while significantly boosting detection efficiency. The findings provide valuable reference for the design and establishment of network intrusion detection models in high-speed railway signal systems.更多还原