基于CNN-ELM的入侵检测

Intrusion detection based on CNN-ELM

针对现有网络入侵检测方法在处理高维度、非线性的海量数据时检测效率低和准确率低的问题,提出一种基于卷积神经网络(CNN)和极限学习机(ELM)的组合式入侵检测方法(CNN-ELM)。通过数值化映射、归一化及维度重组对原始网络数据进行预处理,利用CNN网络自动提取原始网络数据的深层特征,将ELM作为分类器对提取到的特征进行入侵检测分类。采用NSL-KDD数据集对CNN-ELM进行仿真实验,实验结果表明,与SVM、CNN及ELM方法相比,CNN-ELM能有效提高入侵检测的准确率,具有较好的泛化能力和实时性。

Aiming at the low detection efficiency and accuracy of existing network intrusion detection methods in dealing with high-dimensional and non-linear massive data,a combined intrusion detection method(CNN-ELM)based on convolutional neural network(CNN)and extreme learning machine(ELM)was proposed.The original network data were preprocessed by numerical mapping,normalization and dimension reorganization.The deep features of the original network data were automatically extracted through CNN network.ELM was used as classifier to classify the extracted features.Using NSL-KDD data set to simulate CNN-ELM,the results show that CNN-ELM can effectively improve the accuracy of intrusion detection compared with SVM,CNN and ELM methods,and it has better generalization ability and real-time performance.