A Framework for Malicious Domain Names Detection using Feature Selection and Majority Voting Approach

Abstract

As cyber attacks become more sophisticated, identifying and mitigating bad domainnames has become critical to assuring the security of online environments. This paperpresents a framework for detecting malicious domain names using a feature selectionstrategy and a majority vote method. The suggested methodology begins with theextraction of important features from domain names and their related characteristics,followed by a rigorous feature selection procedure to determine the most discriminatingattributes. To accomplish feature selection, a variety of feature selection techniques areused, including chi-square statistics, information gain, gain ratio, and correlation-basedfeature selection, to analyse the value of each characteristic in distinguishing benign andmalicious domain names. In addition, a majority voting strategy is utilised to improvethe detection system’s overall accuracy and reliability by combining the predictions ofdifferent classifiers such as AdaBoost, logistic regression, k-nearest neighbours, naivebayes, and multilayer perceptron. The ensemble of classifiers is trained on the idealfeatures, yielding a complete and robust model capable of accurately recognising mali-cious domain names while minimising false positives. The proposed approach is evalu-ated against real-world examples of harmful domain names. The suggested frameworkemploying Chi-square feature selection and majority voting detects malicious domainnames with an accuracy of 99.44%, precision of 99.44%, recall of 99.44%, and f-measureof 99.44%. The use of feature selection and a majority voting technique improves thesystem’s adaptability and resilience in the face emerging cyber threats.