Performance Evaluation of Machine Learning Classifiers in Malware Detection

Abstract

Nowadays to gain illegal access of android devices or to cause harm to the system, attackers build many malicious software’s. Theses malicious software’s are known as malware. Once device is affected by malware, its performance degrades and more to that there is a risk that your data may be misused by attackers. Over the period of time these malwares have also evolved themselves and detecting a new & generic kind of malwares using conventional techniques is cumbersome and ineffective also. Therefore, it is a need of an hour to make use of some latest approach for detecting malware efficiently. Use of machine learning based techniques can be effective in this purpose. Effectiveness of various machine learning algorithms can be checked by evaluating their performance through certain experiment. In this paper performance of 10 different machine learning classifiers is evaluated on a kaggle dataset containing 15036 malware and benign applications. All the classifiers are evaluated using parameters like Accuracy, AUC, FPR and FNR.