Saldırı Tespiti için Ateş Böceği Algoritması Tabanlı Özellik Seçim Yöntemi ve Yapay Bağışıklık Sistemi

Abstract

Intrusion detection systems generally produce high dimensional data in network-based computer systems. It is required to analyze this data effectively and create a successful model by selecting the important features to save only the meaningful data and protect the system against suspicious behaviors and attacks that can occur in a system. Firefly Algorithm (FFA) is one of the most promising meta-heuristic methods which can be used to select important features from big data. In this paper, a modified Firefly Algorithm-based feature selection method is proposed. The traditional Firefly Algorithm is improved by using the K-Nearest Neighborhood (K-NN) classifier and an additional feature selection step. The proposed method is tested on 4 different datasets of various types of attacks. Three different sub-feature sets are obtained for each dataset and the classification performances are compared. Artificial Immune System (AIS) method is also implemented to generate artificial data for the datasets that have an insufficient number of data. This study shows that the proposed Firefly Algorithm performs successfully to decrease the dimension of data by selecting the features according to the obtained accuracy rates of the K-NN method. Memory usage is dramatically decreased over 50% by reducing the dimension with the proposed FFA. The obtained results indicate that this method both saves time and memory usage.

An ensemble filter-based heuristic approach for cancerous gene expression classification

“…Based on accuracy, the proposed approach performs better than others in the majority of the cases. Considering the same metric, the methods Rani et al [47][48][49][50][51][52][53][54][55][56][57] and Ghosh et al [35] perform better in two cases each thus becomes the second best solution. The three methods, i.e., present proposal, [57], and [35] perform close to each other.…”

“…Before applying their solution, mRMR is employed to reduce the initial features. This helps in reducing the search spaces as well [45][46][47][48][49][50][51][52][53][54] Xue et al [45] present an algorithm called NSGA-III based on three objectives for feature selection. Their solution selects reliable features from an incomplete datasets.…”

“…The performance of their work is evaluated through four sets of medical data from the UCI database. The authors in [49] propose a modified version of firefly algorithm to select important features from the massive datasets generated by the intrusion detection system. They use k-NN method and addition of extra feature selection to improve the traditional firefly algorithm.…”