Key points are not available for this paper at this time.
The Artificial Neural Network (ANN) enables systems to think and act intelligently. In recent years, neural networks are applied in security for such tasks as attack classification, prevention, attack detection… Therefore, there are several researches in this area, particularly in Intrusion Detection System (IDS) which are based on neural network. Most of these IDSs are developed by using Neural Network for Pattern Recognition and are alimented by KDD data. However, some KDD-features have eighter no role, or a minimum impact in attack detection. Consequently, the large number of parameters that must be selected to produce a neural network detecting attacks must be optimized in order to decrease error and increase the IDSs performance. First of all, we have taken some and all of the basic attributes to aliment the network's input and to verify the dependence between these parameters and attacks. Then, we have added the parameters relating to content and time-based ones in order to demonstrate their utility and performance and also to present in which case they are crucial. It must be noted that we have employed MATLAB tool to put into practice these case studies. The analysis is realized in terms of two metrics which are the True Positive Rate (TPR) and the False Positive Rate (FPR).
Farissi et al. (Sat,) studied this question.