Wavelet neural network optimization applied to intrusion detection

Abstract

The wavelet neural network combines wavelet transform and neural network advantages, a strong nonlinear mapping ability and adaptive, self learning, particularly suitable for intrusion detection systems. Wavelet neural network is easy to fall into local minima value, having slow convergence weakness. In this regard, we introduce the genetic algorithm to optimize neural network generating the initial weights and threshold value to determine a better search space, thereby overcoming the neural network easy to fall into local minima shortcomings; identified in the genetic algorithm the search space for fast training of the network, wavelet neural network to solve the traditional slow convergence problems. Simulations show that the method is feasible, the neural network approximation ability and generalization ability has been significantly increased.