Risk Analysis of Android Applications using Static Permissions and Convolutional Neural Network
Kshamta Chauhan, Ekta Gandotra · 2023
Today, Android is the most popular mobile platform for users, vendors, and developers. As the number of Android applications grows, the risk of malware on these devices also increases. Mobile apps do collect a tonne of information about users and their devices, however, not all of it is technically necessary for the operation of the app (e.g., user profiles and habits). App developers are required by privacy laws to disclose to consumers the data usage policies they have implemented (such as what data is gathered and why). Thus, they end up installing apps without realising how that may affect their privacy. In this paper, static permissions are used to detect the risk of the applications. A convolutional neural network is used to predict the probability of the benign and malware classes. The risk of the application is categorised into four parts: no risk, low risk, medium risk, and high risk. Additionally, a graphical user interface is created, upon which the Android app may be uploaded in order to determine its risk using the proposed approach. After the experimental results, the accuracy of the proposed model is 97.02%.