Research On Intrusion Detection Based On Deep Confidence Network

As an proactive security precaution technology,intrusion detection has been applied to network security for a long time.However,with the development of the Internet and the continuous deepening of the network application,in view of the constant changes in the number and the technical level of the attack and intrusion of the network,the traditional misuse based detection technology and the abnormal detection technology can not meet the requirements of the existing network security based on the new type and multi concurrent attack.As the advanced technology of machine learning and artificial intelligence,deep learning has made great achievements in speech recognition,computer vision,large data processing and so on.It also provides a new way of thinking for solving the current intrusion detection problem.Based on the research of traditional intrusion detection technology and deep confidence network under deep learning method,this paper presents an intrusion detection technology based on deep confidence network.According to the disadvantages of intrusion detection data centralization,the data is sampled,and the data after sampling is normalized in non [0,1] interval.In the process of updating the parameters of the depth confidence network,the variable learning rate algorithm with batch gradient descent is adopted,and the updating process of the parameters is added quickly,at the same time in each batch.In training data,the division of fewer categories labels is added to improve accuracy.Finally,the method proposed in this paper is tested on the NSL-KDD dataset.The experiments first determine the best unsupervised learning and the number of supervised learning iterations,and the network structure of the deep confidence network,and then compare with several common intrusion detection methods,such as KNN,SVM and BP network,when the parameters are optimal.Finally,compared with the unimproved DBN.The experimental results show that the method proposed in this paper has a better effect on accuracy,false alarm rate and false rate than other methods,which proves the effectiveness of the proposed method.