Research On Intrusion Prevention Technology Based On Deep Learning And Semi-supervised Clustering

Network security has become a focus of network development for people.Traditional network security technology has firewall, intrusion detection technology and so on, however,the current network environment is so complexly that the traditional network security technology can not meet the needs of network security.Intrusion prevention system based on Intrusion detection system has the advantages of both the firewall technology and intrusion detection technology,which can find not only the traditional intrusion behavior, but also can judge the implicit attack behavior,and take protective measures in time, improve the performance of network security effectively.Intrusion detection algorithm based on distance and probability can find some simple intrusion behavior based on distance and probability usually,but it is difficult to find out the relationship between the characteristics of elements, and is hard to work for the attack means with implicit behavior. Deep learning simulated the human thinking mode, gradually extracted the abstract features, and those features will be used for classification. The mind of Deep learning algorithm is multi-level learning to extract feature, and find the intrinsic link between the data through multiple levels of feature extraction and improve detection rate.Semi-supervised learning uses a small number of labeled data and a large number of unlabeled data to train, it reduces the requirements of the sample. This paper used semi-supervised learning in detection algorithm and proposes an intrusion detection algorithm based on deep learning and semi-supervised clustering. The intrusion detection algorithm based on deep learning and semi-supervised clustering is an improvement to the intrusion detection algorithm based on Shallow learning. In order to train the intrusion detection algorithm based on the Shallow learning, it is necessary to use BP algorithm to carry out the repeated experiments and a lot of labeled samples, and it is easy to generate gradient dispersion in the case of more hidden layers. The intrusion detection algorithm based on deep learning and semi-supervised clustering used sparse auto-encoder to train the hidden layer, and use level greedy algorithm to solve the problem of gradient dispersion. The sparsity used in the algorithm can restrict the activation of the hidden layer units, and caneffectively deal with the problem that the hidden layer units are hard to be determined.The algorithm combined semi-supervised clustering and the deep learning based on sparse auto-encoder, and can train out algorithm parameters on the basis of a certain amount of labeled data efficiently, which has the advantage of high detection rate in the case of using a small amount of labeled data.This paper selects some representative data to test based on the actual situation of the network, and the result are compared with intrusion detection algorithm based on the K-means algorithm, C-means algorithm, Shallow learning algorithm and other intrusion detection algorithms. The experimental results show that the intrusion detection algorithm based on deep learning and semi-supervised clustering can effectively improve the detection efficiency, and overcome influence of initialization and serious noise and other problems in traditional clustering based intrusion detection algorithm. This paper studies the intrusion prevention system based on deep learning and semi-supervised clustering, introduces the working principle of the system, and gives the way of deployment of the system at last.