Hybrid Neural Network Applied Research In Intrusion Detection

With the development of network, the hijack intrudes our host more complicated and covertly. As a result, the traditional technology is discouraged to detect these intrusions. In order to improve the performance of intrusion detection system (IDS), some researchers pay their attention to neural network. In this paper, the achievement of applying neural network in IDS is summarized, the advantages and drawbacks of applying BP and ART-2 in IDS are pointed out, and the solutions to the problem are given out. In the last chapter, a novel IDS named BA-NIDS is implemented based on hybrid neural network. The validity of using hybrid neural network in IDS is certified by experiment.Main contributions of this dissertation are listed as following:Firstly, a review of IDS is presented, and the advantage and drawback of using BP in intrusion detection system are indicated. A novel method which is building different BP based on the different intrusion type is proposed, and experiments are used to validate this model's performanc.Secondly, with the aim of making up the drawback of normal ART-2 suffered, a novel ART-2 neural network model with more vigorous vigilance test criterion is proposed. Then simulation experiments which are based on the KDDCUP 99 data set are used to validat the validation of applying this novel ART-2 on intrusion detection.Thirdly, BP is compared with ART-2 when used in intrusion detection and a novel IDS model which is blending the BP and ART-2 in IDS is put forward. BP is applied to detect specifically intrusion while ART-2 was used to detect unknown intrusion.Finally, a new IDS named BA-NIDS is implemented based on hybrid neural network, and the characters and functions of different components are introduced. The experiment results show that the BA-NIDS is validity and using the hybrid neural network in IDS is feasible.