Please use this identifier to cite or link to this item:
|Title||Wireless Intrusion Detection Enhancement using Backpropagate Neural Network|
IEEE 802.11 wireless networks (WLANs) through its evolution stages are vulnerable for service availability security issues, Denial-of-Service (DoS) attacks are the most danger for the availability of the WLANs. Many of the existing wireless Intrusion detection Systems (WIDSs) have been studied to find a way to enhance the WIDS’ performance. The MAC frame, as a part of the 802.11 wireless frames, contains useful features in detecting the DoS attacks. In this dissertation a traffic has been generated on Infrastructure WLAN by using four different DoS attacks (Airflood, Channel Switch, Quiet and TKIP Cryptographic) to create a suitable data set with two different classes (Normal and Attack) for the Backpropagate Neural Network (BNN) which will be used as a model for anomaly-based WIDS. Four different goals experiments have been done to: measure the performance of the BNN’s model; ranking the data set features by using three different ranking algorithms to find the optimal features set; finding the best BNN’s architecture for both BNN’s models, the one with all features and with optimal features sets; and finally the results of the experiments will be confirmed by using another tool. The main used tool is RapidMiner while the confirmation tool is MATLAB. The experimental results show that the accuracy of the BNN’s models is too closed to 100% and the False Negative/False Positive rates are too small.
|Publisher||the islamic university|
|Files in this item|