On the Optimality of the Simple Bayesian Classifier under Zero-One Loss
Machine Learning - Special issue on learning with probabilistic representations
Feature subset selection bias for classification learning
ICML '06 Proceedings of the 23rd international conference on Machine learning
A fuzzy-genetic approach to network intrusion detection
Proceedings of the 10th annual conference companion on Genetic and evolutionary computation
Mining SQL injection and cross site scripting vulnerabilities using hybrid program analysis
Proceedings of the 2013 International Conference on Software Engineering
| Hi-index | 0.00 |
Network Intrusion detection systems have become unavoidable with the phenomenal rise in internet based security threats. Data mining technique based Intrusion Detection System, have the added advantage of processing large amount of data speedily. However, success rate is dependent on selecting the optimal set of features here. Given an optimal set of features and a good training data set, Bayesian classifier is known for its simplicity and high accuracy. On the other hand, clustering techniques have the flexibility to detect novel attacks even when training set is not present. Therefore, combining the results of both classification and clustering techniques can improve the performance of Intrusion Detection systems greatly. Our project aims at building flexible Intrusion Detection system by combining the advantages of Bayesian classifier and the genetic clustering algorithm. It was tested with KDD Cup 1999 dataset by supplying it with a good training set and a minimal one. In the first case, it produced excellent results, while in the second case it gave consistent performance.