An extrinsic random-based ensemble approach for android malware detection. Issue 4 (2nd October 2021)

• Record Type: Journal Article
• Authors: Potha, Nektaria; Kouliaridis, V.; Kambourakis, G.
• Is Part Of: Connection science. Volume 33:Issue 4(2021)
• Page Start: 1077
• Page End: 1093
• Publication Date: 2021-10-02
• Subjects: Mobile malware -- malware analysis -- machine learning -- ensemble learning -- android security; Neural computers -- Periodicals; Artificial intelligence -- Periodicals; Cognitive science -- Periodicals; Connectionism -- Periodicals; 006.3
• DOI: 10.1080/09540091.2020.1853056
• Languages: English
• ISSNs: 0954-0091

Abstract

Abstract : Malware detection is a fundamental task and associated with significant applications in humanities, cybersecurity, and social media analytics. In some of the relevant studies, there is substantial evidence that heterogeneous ensembles can provide very reliable solutions, better than any individual verification model. However, so far, there is no systematic study of examining the application of ensemble methods in this task. This paper introduces a sophisticated Extrinsic Random-based Ensemble (ERBE) method where in a predetermined set of repetitions, a subset of external instances (either malware or benign) as well as classification features are randomly selected, and an aggregation function is adopted to combine the output of all base models for each test case separately. By utilising static analysis only, we demonstrate that the proposed method is capable of taking advantage of the availability of multiple external instances of different size and genre. The experimental results in AndroZoo benchmark corpora verify the suitability of a random-based heterogeneous ensemble for this task and exhibit the effectiveness of our method, in some cases improving the hitherto best reported results by more than 5%.