Supervised Learning Methods Comparison for Android Malware Detection Based on System Calls Referring to ARM (32-bit/EABI) Table
Journal of Information Technology and Cyber Security · 2024
DOI: 10.30996/jitcs.10511 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Information Technology and Cyber Security, 2024. Verfasst von Rinanza Zulmy Alhamri und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Android malware detection research is a topic that is still being developed. From all the detection techniques developed, dynamic analysis methods have become interesting because they trace the suspect application system calls. Based on the system calls, by utilizing machine learning, the suspect application can be classified as malware or benign. Comparing the machine learning methods is im-portant to determine what method is best to support malware detection. This article aims to explain more clearly and simply the way to conduct Android malware detection based on system calls step by step using classification. Furthermore, it presents the system calls sequence conversion referring to the arm(32-bit/EABI) table, which has 398 system calls (0-397) as features. It will provide a compari-son of several supervised machine-learning methods for classifying Android applications. This initial research is part of the other research that has the purpose of developing a malware detection system based on an Android application. This research can be used to develop the best machine learning to classify malware applications using a Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbour (KNN), and Naive Bayes (NB). The result can be concluded that the KNN method has the lowest performance in detecting Android malware apps, with an accuracy of only 0.50. In comparison, the NB method has an accuracy of only 0,69. SVM and DT models have similar accuracy and recall results of 0.79 and 0.75, respectively, but DT obtained higher precision and scores of 0.83 and 0.76, respectively. Although in this study, the classification performance of DT is better than SVM, based on comparison with the results of previous research, SVM is a suitable method for Android malware de-tection based on system calls. It is proven by the results of research comparisons that the SVM method is always the method with the highest accuracy score among other methods. For the next research, the SVM method can be used to develop a malware detection system for Android applications.
Zitieren
APA
Alhamri, R. Z., Cinderatama, T. A., Eliyen, K., & Izzah, A. (2024). Supervised Learning Methods Comparison for Android Malware Detection Based on System Calls Referring to ARM (32-bit/EABI) Table. Journal of Information Technology and Cyber Security. https://doi.org/10.30996/jitcs.10511
Vancouver
Alhamri RZ, Cinderatama TA, Eliyen K, Izzah A. Supervised Learning Methods Comparison for Android Malware Detection Based on System Calls Referring to ARM (32-bit/EABI) Table. Journal of Information Technology and Cyber Security. 2024. doi:10.30996/jitcs.10511
BibTeX
@article{Alhamri2024,
author = {Alhamri, Rinanza Zulmy and Cinderatama, Toga Aldila and Eliyen, Kunti and Izzah, Abidatul},
title = {Supervised Learning Methods Comparison for Android Malware Detection Based on System Calls Referring to ARM (32-bit/EABI) Table},
journal = {Journal of Information Technology and Cyber Security},
year = {2024},
doi = {10.30996/jitcs.10511}
}
EUVIMED