Enhancing Android Platform Security: Investigating Malware Patterns with Sufficient Input Subset
Farrakh Nazir, Neeli Khan, Muhammad Usman Shahid Khan, Ahmad Fayyaz · 2023
Smartphones have become an integral part of our reality, while there are various machine learning algorithms available for detecting malware in android applications, they often lack explain-ability due to the opaque nature of "Black Box" decision-making. To address explain-ability, a combination of the Sufficient Input Subset (SIS) technique and convolutional neural network (CNN) is used. SIS identifies minimal feature subsets that are sufficient for reaching the same verdict based solely on their observed values. The proposed approach has several benefits, including reduced computational complexity and increased speed of detection, as it requires fewer features to decide which part of APK contains the malware.