Detection Success Assessment of Machine Learning Algorithms Through Manifest File Permissions Demanded by Malicious Android Wares


Aytac Doganay H. A., BÜLBÜL H. İ.

22nd IEEE International Conference on Machine Learning and Applications, ICMLA 2023, Florida, United States Of America, 15 - 17 December 2023, pp.1684-1686 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/icmla58977.2023.00254
  • City: Florida
  • Country: United States Of America
  • Page Numbers: pp.1684-1686
  • Keywords: android, dataset, drebin, machine learning, malicious, weka
  • Ankara University Affiliated: No

Abstract

Due to the open source and widespread use of the Android operating system, malicious threats to devices with this operating system are quite high. In terms of these features, there are many models and studies developed using machine learning and deep learning methods for Android malware detection in the literature. Many data sets containing static and dynamic features of malicious and benign Android software are used in the development of learning methods. Although there are a wide variety of datasets, one of the richest datasets is the Drebin dataset. Due to the need to detect the malware quickly, the Drebin dataset was cleaned and reduced to manifest file permissions, and as a result of Random Forest, NaiveBayes, J48 and AdaBoost algorithms applied using Weka software on this dataset in the literature, it was seen that the Random Forest algorithm achieved the most successful result.