Intrusion Detection System Based On Artificial Neural Network

With the development at full speed of global informationization,the computer onlinesecurity question is outstanding day by day. The danger that the hacker invades,infromationreveals and the virus overflows has caused the great attention of the countries all over the world,so network information security has already become a fatal problem to settled. The traditionalfirewall,router and the techniques of identity authentication and data encryption can not fitcurrent network environment. Intrusion Detection System(IDS)is the research hotspot in thefield of Network Security now,which plays an important role in safeguarding network securityand people pay more attention to.Though IDS can detect, react and protect the information system in real time in theory, ithas its own evolution like any other systems. IDS products have some limitation and fragility atpresent, and may receive various kinds of attacks standing in the breach. With the popularizationand application of Internet, how to effectively find, follow and deal with the potential hazard ofsafety in time in the environment of Internet has proposed higher demand to IDS,and traditionalIDS can not fit current network environment.At home and abroad, we generally adopt the intelligent technologies to solve theseproblems in IDS at present. Under this background,we bring forward the idea of applyingArtifical Neural Networks(ANN)in it. It can improve the efficiency and strengthen the learningability of the system if we apply the multi--layer Back Propagation Neural Network which beresearched and employed most extensively to IDS Especially.This paper does careful analysis and research in the Intrusion Detection technique and ANNtechnology,and designs a Intrusion Detection System model based on ANN,then realized it. Themain researches in this thesis are as follows:1. Further investigate and analyze the Intrusion Detection technique and artificial neuralnetwork technologyThis paper opens with some elemental conceptions and theories of IDS,then analyses themerit and flaw of intrusion detection system existing,points out the shortcoming and challengefaced,and concludes its developing direction. Then studies upon on neural network,and pointsout the wide research prospect in the application of ANN in IDS. The author analyses the model,topology structure and learning rule of ANN further.2. Propose the improved algorithm of the BP study algorithm in the multi-layer feedforwardnetworkWe carry on more detailed discussion about multi-layer feedforward network and BP Studyalgorithms,and improve BP algorithm for its low learning efficiency and slow restraint speed.We adopt additional momentum law and adaptive learning speed to raise the performance of BPalgorithm.3. Put forward a detailed design scheme of intrusion-detection model based on artificialneural network and realized itWe research and analyze the application of ANN in the IDS,and point out the system willplay a much role in the theory and practice if it can be designed and implemented. On thisfoundation and in the reference of CIDF standard,we put forward a detailed design scheme ofintrusion-detection model based on ANN. Great emphasis was put in key modules—the neuralnetwork classified module,then explain the Operation principle of model in detail,and givedetailed introductions to every module in the model. Lastly according experimental result throughtraining and intrusion procedure,we get a fairly analysis,which indicates the neural network hasa very great advantage in intrusion detection.We have also found some problem during the study,which we will try to research and solvein the future study,and we will continue trying the more effective way to applying NeuralNetwork in the Intrusion Detection Systems.