Intrusion Detection Based On Neural Computation And Evolutionary Network

Intrusion detection is a disquisitive hotspot in network security. It analyses the collected data to detect complicated and dormant attacks. Clustering which employs unsupervised learning algorithm is a pivotal technique of Data Mining. And clustering is an important technique of data analyse. Intrusion detection based on clustering can not only conquer the shortages of the traditional algorithms but also improve the performance of Intrusion detection system (IDS) effectively.The advantages and disadvantages of the existing IDS are analysed in this paper. In order to solve their problem, to the viewpoint of intelligent complementary fusion, a clustering algorithm which based on Neural Computation and evolutionary network is employed to our disquisition of IDS. The research works are as follows:1. The actuality that some existing clustering procedures applied to intrusion detection is analysed. And the advantages and disadvantages of the methods are summarized.2. Artificial immune network is made as the emphasis of our investigation. Immune network clustering is independent of data distribution and transcendental knowledge. And it can find the solution to the demand of clusters number beforehand. Both intrusion and unknown attacks can be detected by using it in intrusion detection.3. The algorithm of intrusion detection based on ART and artificial immune network is proposed in this paper. That is: A large amount of data for intrusion detection is pretreated by ART network, and then the outputs of ART are considered as initial antibodies to train an immune network. Last minimal spanning tree is employed to perform clustering analysis and obtain characterization of normal data and anormal data. The computer simulations on the KDD CUP99 dataset show that both the convergence speed and the precision of this immune network are better than the one without the pretreatment of ART. And the method can deal with mass unlabeled data to distinguish between normal and anomaly and to detect unknown attacks.