Performance Comparison of Intrusion Detection System Using Various Techniques - A Review

Nowadays, the security has become a critical part of any organization or industry information systems. The Intrusion Detection System is an effective method to deal with the new kind of threats such as DoS, Porbe, R2L and U2R. In this paper, we analyze the various approaches such as Hidden Semi Markov Model, Conditional Random Fields and Layered Approach, Bayesian classification, Data Mining techniques, Clustering Algorithms such as K-Means and Fuzzy c-Means, Back Propagation Neural Network, SOM Neural Network, Rough Set Neural Network Algorithm, Genetic Algorithm, Pattern Matching, Principle Component Analysis, Linear Discriminant Analysis, Independent Component Analysis, Multivariate Statistical Analysis, SOM/PSO algorithm etc. The performance is measured for two different datasets using various approaches. The datasets are trained and tested for identifying the new attacks that will affect the hosts or networks. The well known KDD Cup 1999 or DARPA 1999 dataset has been used to improve the accuracy and performance.

The four groups of attacks are identified as Probe, DoS, U2R and R2L. The dataset used for training set is 494,021 and testing set is 311,028. The aim is to improve the detection rate and performance of the proposed system.