Intrusion Detection Based On Immune Wavelet Network

The problem of intrusion detection is a division between the normal behaviors and the abnormal behaviors. In this regard we can consider it a problem of pattern recognition (ie. the problem of classification).The main research work in this paper are as following:1. An IDS model based on wavelet networks is proposed. This model combines the advantages of neural networks(the ability of self-learning and associative memory) and wavelet analysis(the ability of the best function approximation).2. Because the efficiency and accuracy of solving the real problem mostly depends on the appropriate learning algorithm, the immune algorithm which replaced the learning algorithm based on the gradient in our IDS model is proposed. It overcomes the shortages such as low speed of convergence easily running into the local optimization etc.3. Immune clone selection algorithm is proposed in this paper. Through constructing the clone operator suitable for the given problem and controlling the size of clone according to the affinity between the antibody and antigen dynamically it can realize the three processes including exploration, guiding and convergency-inducing very well.4The hierarchical structure coding scheme is used in our EDS model firstly which really make the topologic construction and the parameters of the wavelet network evolve at the same time.