基于卷积神经网络的入侵检测研究

Research on Intrusion Detection Based on Convolutional Neural Network

传统机器学习方法在面对庞大、多维的网上数据时,无法满足入侵检测的准确性和实时性的要求,而大规模的数据刚好为深度学习提供了训练数据,深度学习可以直接从数据中获取高级特征.为此,本文提出一种基于卷积神经网络(Convolutional Neural Network,CNN)的入侵检测方法,首先通过数据的预处理去除了5个对结果影响较小的特征,使得预处理结束后数据集的维度刚好满足卷积神经网络的输入要求,然后通过遗传算法(Genetic Algorithm,GA)对卷积神经网络的权值进行优化,解决了参数调优难的问题.将训练好的模型在KDDCUP99数据集上进行验证,实验结果表明本模型的收敛速度较快,准确率高于98.5%,有重要的研究价值.

Traditional machine learning methods cannot meet the accuracy and real-time requirements of intrusion detection when faced with the huge and multi-dimensional online data.The large-scale data provides training data for deep learning,and deep learning can directly obtain advanced features from the data.To this end,an intrusion detection method based on Convolutional Neural Network(CNN)is proposed.First,five features that have little impact on the result are removed through data preprocessing,so that the dimensionality of the data set after the preprocessing is completed just meet the input requirements of the convolutional neural network.After that,the weights of the convolutional neural network are optimized by genetic algorithm(GA),which solves the problem of difficult parameter tuning.The trained model is verified on the KDDCUP99 data set.The experimental results show that the convergence speed of this model is faster and the accuracy rate is higher than 98.5%,which has important research value.