Title Of Paper:
 
 
The Combined Method of Classification for Improvement of Intrusion Detection System in Computer Networks
Author's Name :  Ehsan vahid, Abdol Hamid Fetanat, KarimAnsariasl
KeyWords:   intrusion detection system, Adaboost, kDD-Cup99.
Pages:  75 -88
Volume: 3
Issue: 3
Year: 2015
Abstract:
 

 

 

 

The objective of an intrusion detection system, is classifying activities into two major categories including: normal and attack. The objectiveof this project is using one ofthe algorithm classification methods called Adaboost through applying the necessary changes to increase the efficiency of the algorithm in order to detect attacks. The first change was the weight of initial sample ofAdaboost algorithm. According to the Asymmetric distribution of data it causes some classes not to teach properly. The first change was the algorithm Adaboostin order to initialize the weight of the samples; rather instead of bearing the equal weight into all of the samples the probability of the primary selection of all the samples for learning the algorithms is equal together. The initial weight of the samples was allocated based on the number of them in each class that it solves the problem related to attacks of asymmetric data. Secondly, change is that the combine of the Adaboost algorithm with harmony search algorithm and generate the optimal coefficients for decision Adaboost proposed to detect attacks ; nevertheless, because the harmony search algorithm objective function was affected, the optimum coefficients were affected by asymmetric data, so thatinstead of an array of factors to decide, the coefficient matrix was created forthe number of rows in the matrix which equals to the number of class in the data collection and the number of columns equals to the number of algorithms used in Adaboost. The proposed method was assessed by intrusion detection data sets kDD-Cup99. Compared with other methods, performance of the proposed method is confirmed.

Full Text: