Analysis of KDD '99 Intrusion Detection Dataset for Selection of Relevance Features

摘要: The rapid development of business and other transaction systems over the Internet makes computer security a critical issue. In recent times, data mining machine learning have been subjected to extensive research in intrusion detection with emphasis on improving accuracy classifier. But selecting important features from input lead simplification problem, faster more accurate rates. this paper, we presented relevance each feature KDD '99 dataset class. Rough set degree dependency ratio class were employed determine most discriminating for Empirical results show that seven not relevant any network-based. former operates information collected within an individual system latter collect raw networks packets as source network analyze signs intrusions. two different techniques IDS search attack patterns are Misuse Anomaly. find known signatures monitored resources. Anomaly attacks by detecting changes pattern utilization or bahaviour system. Majority currently use either rule-based expert-system based. Their strengths depend largely ability personnel develops them. can only detect types is prone generation false positive alarms. This leads intelligence technique mining/machine alternative expensive strenuous human input. These automatically learn extract useful reference normal/attack traffic behaviour profile existing subsequent classification