Title: Improving android malware detection: hybrid models integrating SGDC with SHO and GNDO optimisation

Authors: Peng Chen; Yan He

Addresses: Chongqing Medical and Pharmaceutical College, Chongqing, China ' Chongqing Medical and Pharmaceutical College, Chongqing, China

Abstract: Android app development is evolving rapidly, but Android malware is dangerous. Many approaches for malware detection have been proposed and researched in studies, with one such effective tool being Machine Learning (ML). In this study, Stochastic Gradient Descent Classification (SGDC) was utilised for Android malware prediction, supplemented with Sea Horse Optimisation (SHO) and Generalised Normal Distribution Optimisation (GNDO) for increased predictive performance. Hybrid models were developed by combining these optimisation algorithms: SGDC with SHO (SGSH) and SGDC with GNDO (SGGN). SGSH achieved a testing accuracy of 0.990, with SGGN following closely with 0.972 accuracy. For precision, SGSH again performed best at 0.991, with SGDC reporting a minimum of 0.952.

Keywords: android malware; machine learning; stochastic gradient descent classification; sea horse optimisation; generalised normal distribution optimisation.

DOI: 10.1504/IJIPT.2025.147732

International Journal of Internet Protocol Technology, 2025 Vol.18 No.2, pp.90 - 98

Received: 27 Feb 2025
Accepted: 10 Apr 2025
Published online: 28 Jul 2025 *

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