A hybrid intelligent agent based intrusion detection system

In recent years, the number of attacks on data sent through computer systems or networks has rapidly increased. Therefore, interest in developing Intrusion Detection System (IDS) has increased among various researchers in the field of network security. Due to this, many IDS have been developed by these researches. However, with all the existing IDS, the amount of security provided is not sufficient to prevent the intruders and hence the network is still under threats. Therefore, it is necessary to propose and implement new types of security algorithms based on Artificial Intelligence Techniques so that it is possible to enhance the security of data through intelligent mechanisms. This paper proposes a hybrid intelligent agent based IDS by introducing three different types of intelligent agents namely a feature selection agent to select the required features efficiently using rough sets, a validation agent to validate the selected features and to pass on the data to the classifiers C4.5 and SVM and finally a decision agent for making the final decision. This decision agent has been incorporated as a subsystem of into a decision manager, which is used to pick up all the classes, which are classified as normal as well as abnormal, and to analyze and detect the intruders by the above-mentioned three classifiers. These classified results are passed on to an ensemble sub module of the decision manager for making a final decision on intrusions. The ensemble sub module analyses the differences in misclassification and improves the overall accuracy. The experimental results show that the proposed hybrid intelligent agent based model improves the overall detection accuracy and minimizes the computational complexity of classification due to feature selection. © 2005 by Binary Information Press.