A New Networks Intrusion Detection Architecture based on Neural Networks
Networks intrusion detection systems allow to detect attacks which cannot be detected by firewalls. The false positive and false negative problem tend to make IDS inefficient. To improve those systems’ performances, it is necessary to select the most relevant that will lead to characterize a normal profile or an attack. We have proposed in this paper a new intrusion detection system architecture and a scheme to flexibly select groups of attributes using neural networks in order to improve results that we have got with our architecture. The selection approach is based on a contribution criteria that we have defined in function of precision measures of type HVS (Heuristic for Variable Selection).The selected subset depends on a threshold that we make vary in function of a defined criteria. He have done a comparative study of this approach and the one without attributes selection. A comparative study has also been done with others works. The NSL-KDD dataset has been used to train, teste and evaluate our scheme. Our Works shows satisfactory results.
