Intrusion Detection Using BP Optimized by PSO
In this paper a model of intrusion detection is proposed, which assimilates the superiority of global search optimal solution of Particle Swarm Optimization (PSO) algorithm and the advantage of gradient descent local search of Back Propagation (BP) algorithm. Intrusion detection can be performed after BP neural network is trained on the basis of these optimized parameters and samples. Optimized parameters are found by PSO which can prevent BP neural network from falling into local minimum and improve its convergence rate. The experimental results show that the average detection rate of the intrusion detection algorithm based on PSO-BP is 93.4%, and it can provide intrusion detection service effectively and reliably.