Title: Enhancing malware detection in Android application by incorporating broadcast receivers

Authors: Halil Bisgin; Fadi Mohsen; Vincent Nwobodo; Rachael Havens

Addresses: Department of Computer Science, University of Michigan-Flint, Flint, MI, USA ' Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, University of Groningen, Groningen, The Netherlands ' Financial Industry Regulatory Authority, Rockville, MD, USA ' AVL Test Systems Inc., Plymouth, MI, USA

Abstract: Android's large market share makes it attractive for malicious apps, which continue to exist in online markets despite vetting procedures. To alleviate this issue several malware detection methods have been developed-based primarily on the requested permissions and other components that could be extracted from the AndroidManifest.xml file. However, very few of them referred to or thoroughly analysed Android broadcast receivers (ABR). In this paper, we extensively study ABR for characterising and detecting malicious Android applications. Namely, we investigate their use patterns, study their correlations with permissions, and build prediction models with various scenarios and subset of features. Our findings show that ABR are being heavily utilised by malware apps, and are also strongly correlated with permissions. Additionally, combining ABR with permissions improves the accuracy of our prediction models and an optimal subset (~22%) of this combination achieves the same performance which further makes is less computationally demanding.

Keywords: Android; broadcast receiver; privacy; permissions; support vector machines; SVMs; association rule mining; feature selection.

DOI: 10.1504/IJIPSI.2021.119168

International Journal of Information Privacy, Security and Integrity, 2021 Vol.5 No.1, pp.36 - 68

Received: 20 Jul 2021
Accepted: 28 Jul 2021

Published online: 24 Nov 2021 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article