Intrusion Detection Algorithm of EPSO Combined with BP Neural Network

Aiming at the problems of large number of weights and thresholds of back propagation neural network (BPNN), a novel back propagation neural network integrated with particle swarm optimization based on entropy model (EPSO-BPNN) is proposed to improve the optimization effect of weights and thresholds for intrusion detection algorithm. Firstly, the entropy model was introduced to obtain the search characteristics of particle swarm optimization (PSO) algorithm. Then, the entropy value was used to increase the inertia weight timely. Furthermore, the optimal particle was decoded as the weights and thresholds obtained by particle swarm optimization based on entropy model (EPSO) algorithm. They are further optimized in the training phase. The experimental results show that EPSO-BPNN algorithm has a higher solution accuracy and convergence rate than PSO-BPNN algorithm in entropy value, fitness value and mean square error. The intrusion detection rate of EPSO-BPNN algorithm is 92.90%, and it has better intrusion detection rate on five network data types.