改进的BP神经网络与D-S证据理论融合在入侵检测系统中的应用的中期报告.docx

改进的BP神经网络与D-S证据理论融合在入侵检测系统中的应用的中期报告 【摘要】：针对网络入侵检测技术的发展趋势以及其方法的不断完善与更新，本文提出一种改进的BP神经网络与D-S证据理论结合的入侵检测系统。该系统分为数据预处理模块、特征提取模块、BP神经网络模块和DS证据理论模块四大模块，考虑到多类别的入侵类型，模型中利用OneVsRest策略实现多类别的分类任务。本文采取五大特征选取算法，对NSL-KDD数据集中的所有特征进行筛选和评估，最终将34个特征精简至4个输入层特征和3个输出层特征。同时，该系统提出了一种新的D-S证据理论中的证据模糊化方法。实验结果表明，该系统较传统入侵检测系统有较好的性能表现和更强的鲁棒性。 【关键词】：入侵检测；BP神经网络；D-S证据理论；特征选取；证据模糊化。 【Abstract】：In view of the development trend of network intrusion detection technology and the continuous improvement and updating of its methods, this paper proposes an improved BP neural network and D-S evidence theory combined intrusion detection system. The system is divided into four modules: data preprocessing, feature extraction, BP neural network, and DS evidence theory. Considering the multi-category intrusion types, the model uses the OneVsRest strategy to achieve multi-category classification tasks. This paper adopts five feature selection algorithms to screen and evaluate all features in the NSL-KDD dataset, ultimately reducing 34 features to 4 input layer features and 3 output layer features. At the same time, this system proposes a new evidence fuzzification method in D-S evidence theory. The experimental results show that the system has better performance and stronger robustness than traditional intrusion detection systems. 【Keywords】: intrusion detection; BP neural network; D-S evidence theory; feature selection; evidence fuzzification.