Intrusion Detection System Based On Genetic Algorithm And Wavelet Neural Networks

Wavelet neural network has a very strong nonlinear ability and the self-adaptability, it combines the advantage of wavelet transform with that of neural network, so it can be applied very well in the fields of intrusion detection. But it also has some weakness in practice, such as the slow convergence speed and easy convergence to the local minimum points. Therefore, the genetic algorithm is introduced into the computer system to improve it. We first use the genetic algorithm to optimize the initial weight values of the wavelet neural network to locate a better solution space so as to prevent its tending to fail into the part infinitesimal plot. At the same time, in order to solve the low speed in constringency, an optimized Levenberg-Marquardt(LM) algorithm is also introduced to accurately train the neural network. Both works together as LM-WNN. The simulation results indicate that this combination is feasible, leading to the approximation capability and generalization ability of the network be enhanced.