Intrusion Detection System Based On Artificial Neural Network

When more and more companies shift its core business to the Internet, network security as an unavoidable question present the people. Along with aggressor's knowledge mature day by day, attacked tool and technique is becoming increasingly complex and diverse, traditional static strategy has been unable to meet the security of highly sensitive department needs, the network defense must use a deep and various defenses. As an active network security protection system, IDS not only can detect intrusions from the outside, but also can supervise unauthorized users, has a very broad range of applications.However, there is still widespread defects from the traditional IDS in the detection performance, scalability, distributed, adaptive, self-learning, robustness.This paper establish a simple and effective IDS based on ANN. This system use many advantages of ANN such as self-learning, adaptive,non-linear,parallel processing, robustness and fault-tolerant, improving the system's robustness and adaptability.First, the paper analyzes the header information of data packet for TCP/IP bottom protocol (TCP, IP, UDP, ICMP), and introduce the principle, method of common network attacks in detail.Then, propose an design programmer of overall for intrusion detection system based on artificial neural network, and withdrew 11 various key words by principle of key words select as input vector for artificial neural network.Finally, the paper takes training and testing to the IDS, the results show that the system not only can detect attacks on a high detection rate, but also have a good detection capability on unknown attacks, furthermore it has low rate of misuse warn.