Research On Intrusion Detection Based On BP Neural Network Algorithm

In recent years, network security problem has been received more and more attention,as one of the important technology of network security, intrusion detection technology hasundergone nearly30years of development, but there are still a number of deficienciesfactors (such as bad real-time detection, needmanually update the rule base, etc.), BPneural network is applied to the network intrusion detection and effective response tothese deficiencies, but there is a system false alarm rate, high false negative rate.In this paper, focus on these defects, we use modified particle swarm optimizeralgorithm combined with BP neural network in the intrusion detection system, mainlyfocus on the detection with Dos attacks and Probing attacks, which has achieved a certaineffect in the optimization of the system false negative rate, false alarm rate and currentlyof the convergence rate. The neural network is widely used in various fields and achievedgood results, the research of this article for this attempt, mainly because:(1)Distinguishing with the traditional intrusion detection technology, the neural network isapplied to intrusion detection by training to get the ability to predict.(2) As a newinvasion, the method of re-training of the neural network to respond to the new attacks toensure that the adaptive capacity of the system.(3) Network intrusion detection can alsobe regarded as the data to be detected pattern recognition problem.This paper describes the network attacks and intrusion detection technology and thedevelopment direction of the principle analysis of various attack techniques.With analysis of the BP neural network detailed, a certain degree of optimization onthe defects of the standard BP neural network (such as easy to fall into local minimum,slow convergence speed, etc.), through the gradient descent and additional momentumalgorithm optimization, etc.By the structure of the BP neural network system, build individual and population ofthe genetic algorithm, a detailed analysis of the genetic algorithm is applied to BP neuralnetwork initial weights and threshold optimization problem, with particle swarmoptimizer algorithm process, the BP neural network used in intrusion detection.Introduce a PSO-BP neural network intrusion detection system model, simulationexperiments with the KDDCUP1999data. The experiment results show that the model of intrusion detection detect Dos attack, and Probing scan attack, in terms of theoptimization of the system false negative rate, false alarm rate and convergence rate, notonly achieved certain results, but also has a good variant for known attacks and unknownattacks the ability to detect the particle swarm optimizer algorithm applied to theeffectiveness and feasibility of the intrusion detection system combined with BP neuralnetwork.