Intelligent network-based intrusion detection system (iNIDS)

Abstract

Networks are regarded as one of the biggest advancements in the field of computer science. But they enable outsiders to "intrude" into our information. Intrusions can be in the form of simple eavesdropping, or gaining access to the host itself. Here, intruders are identified using two main methods - signature analysis and anomaly analysis. The proposed method is such that the signature analysis is strengthened by anomaly analysis, which in turn uses some level of intelligence based on the traffic parameters, obtained and processed using neural networks. The initial intelligence is obtained using the KDDCUP99 dataset, which trains a neural network. The neural network will take care of further detections, and it strengthens itself during the run itself. The result obtained suggests that even with minimal initial intelligence, iNIDS can reach accuracy levels of over 70%, and by increasing the initial set a little more, it reaches accuracy levels exceeding 80%. © 2012 Springer-Verlag.