基于VDCNN和LSTM混合模型的入侵检测算法

Research on Intrusion Detection AlgorithmBased on VDCNN and LSTMHybrid Model

为了提高入侵检测系统在复杂数据集下的分类性能,提出一种将超深度卷积网络(very deep convolutional neural networks,VDCNN)和长短时记忆网络(long short-term memory,LSTM)混合模型的入侵检测算法。本模型通过CICFlowMeter工具对CES-CIC-IDS2018数据集进行特征采集,并对采集到的特征进行清洗、转换等预处理;将处理好的数据分别传入VDCNN网络和LSTM网络中获取数据特征和关联特征;最后将两类特征传入融合层以实现在特定维度上的拼接,形成新的数据特征进行分类识别,得出检测结果。使用了多种对比方法进行验证,实验结果表明所提分类模型相较与其他模型有效提高了入侵检测识别的准确率。

A hybrid intrusion detection method of bybird VDCNN and LSTM models,is presented to increase the classification performance of the intrusion detection technology under complex data sets.The feature collection with the CICFlowMeter on the CES-CIC-IDS2018 data set is done with the model;the collected features are cleaned and converted and other pretreatments are made.Then the processed data are transmitted to VDCNN network and LSTM network respectively to obtain data features and association features.Finally,the two types of features are transmitted to the fusion layer to realize the splicing in specific dimensions,the newly formed features are used for classification and recognition.The detection results are obtained.Many kinds of comparison methods are used for validation.The experiment results show that the proposed classification models effectively improve the accuracy of intrusion detection and recognition rate compared with the other model.