Examining Permission Patterns in Android Apps using Kernel Density Estimation
Muhammad Suleman Saleem, Jelena Mišić, Vojislav B. Mišić · 2020 International Conference on Computing, Networking and Communications (ICNC) · 2020
Effective detection of malware apps requires thorough knowledge of behavioral and structural patterns of both benign and malware apps. In this paper, we focus on permissions which allow or restrict access to system services in Android mobile operating system. After a preliminary statistical analysis of different permissions classified into normal, dangerous, and signature ones, we apply Kernel Density Estimation, a nonparametric method for estimating probability distribution of data. Our analysis is based on the distinct permission patterns in both benign and malware apps from a research database containing nearly 120,000 sample apps. The research presented here is the necessary first step towards building a cost-effective malware app detector for the Android system.