Filter Based Data Reduction Technique and Classification for Intrusion Detection System
Keywords:
Intrusion detection, Feature selection, Mutual information, Linear correlation coefficient, Least square support vector machine.Abstract
Redundant and tangential options in information have caused a semi permanent downside in network traffic
classification. These options not solely curtail the method of classification however conjointly forestall a classifier from
creating correct choices, particularly once dealing with huge information. During this paper, we have a tendency to
propose a mutual info primarily based algorithmic rule that analytically selects the best feature for classification. This
mutual info primarily based feature choice algorithmic rule will handle linearly and nonlinearly dependent information
options. Its effectiveness is evaluated within the cases of network intrusion detection. Associate in Nursing Intrusion
Detection System (IDS), named Least sq. Support Vector Machine primarily based IDS (LSSVM-IDS), is constructed
exploitation the options hand-picked by our projected feature choice algorithmic rule. The performance of LSSVM-IDS is
evaluated exploitation 3 intrusion detection analysis datasets, particularly KDD Cup ninety nine, NSL-KDD and urban
center 2006+ dataset. The analysis results show that our feature choice algorithmic rule contributes additional
important options for LSSVM-IDS to realize higher accuracy and lower process price compared with the progressive
strategies.
