The Technology Research Of Network Security Based On Network Traffic Anomaly Detection

With the development of new technologies and the expansion of networked computer systems, sensitive data are under constant threat of attack from hackers. Intrusion detection can recognize the attacks which attempt to happen, are happening or have happened, and it is a kind of active network security protection measure. It has been widely concerned by researchers as an important part in the network security filed. The existing intrusion detection technologies have the deficiency of high false positive rate, higher false negative rate, and poor real-time performance. Especially high detection accuracy is usually based on abundant or self-contained training data.In this dissertation, difficulties that traditional intrusion detection is confronted with in new network environment are analyzed. In order to overcome these difficulties, a new detection model---Neural Network Trained by Swarm Intelligence Model is presented.Firstly, intrusion detection's conception, sorts, characteristics, research content and difficulties confronted by the traditional intrusion detection are analyzed systematically. Then we formulate the principle of some widely known neural network and discuss the concept of Particle Swarm Optimization and Quantum-behaved Particle Swarm Optimization. We show that QPSO-Trained RBF can lead to solution with higher precision and faster convergence speed than PSO-Trained RBF. Neural network's conception, characteristics, structure and the training algorithm of wavelet neural network are analyzed. In turn, we proposed an approach of using QPSO to train Wavelet Neural Network.Then, we use QPSO-Trained WNN to intrusion with Genetic Algorithm (GA) Trained WNN and PSO-WNN also tested for the purposed of performance comparison. The well-known KDD Cup 1999 Intrusion Detection Data Set was used as the experimental data. Experimental result on KDD 99 intrusion detection datasets shows that the accuracy of anomaly detection was enhanced and the false positive rate for normal state in the network anomaly detection was declined.Finally, the modified QPSO algorithm is presented, which is use as the training algorithm of wavelet fuzzy neural network. So the model which is tested on the KDD99 database is built.The work in this paper indicate that QPSO algorithm and modified QPSO algorithm are promising training algorithms for neural network and could generate better performance than other intelligent optimization algorithms such as GA and PSO. It will work well on intrusion detection problem by neural network.