An Artificaial Intelligence Integrated Framework And Methodology For Information Security

Abstract

In computer network, the presence of any malicious node can have a major impact on the entire communication. There are several attacks like Blackhole attack, wormhole at-tack in which a genuine node behaves as malicious node and affects the performance of the network. There have been many researches for detecting the attacks in the network-ing system but regarding identifying and determining the behaviour of node in net-works, there are very few researches. Thus it is essential to determine the normal and malicious behaviour of the node. newlineThe motivation for undertaking this research was the deficiencies encountered in previ-ous researches on detection of malicious node intrusion. A malicious node violates the security principles and affects the network adversely thereby degrading the perfor-mance of the network.The training and testing time involved to build intrusion detec-tion model for malicious attacks is very high in most of the cases. The false alarm rate is also high in many studies. These deficiencies imply that there is need of a secure intelli-gent framework that will effectively identify the malicious traffic and the malicious node and analyse the behavior of the node. newlineThe experiments were performed on real-time standard dataset on weka simulator. The dataset has normal and malicious class both for a large number of traffic records.To re-duce the dimensionality of dataset, the individual feature selection algorithms are per-formed on the dataset and results of classification are recorded. These results have mo-tivated to introduce a novel feature selection algorithm named as Ensemble Feature Se-lection Algorithm. The malicious traffic has been correctly identified with higher accu-racy and low false alarm rate. newlineThe identification of malicious node is determined by applying the same proposed algo-rithm on the dataset and the modified dataset is clustered by using KMeans algorithm. The classification of these clustered results determine the malicious node with the help of the srcip i.e. IP address of the