Classification and security assessment of android apps

Abstract

Abstract Current mobile platforms pose many privacy risks for the users. Android applications (apps) request access to device resources and data, such as storage, GPS location, camera, microphone, SMS, phone identity, and network information. Legitimate mobile apps, advertisements (ads), and malware all require access to mobile resources and data to function properly. Therefore, it is difficult for the user to make informed decisions that effectively balance their privacy and app functionality. This study analyzes the Android application permissions, ad networks and the impact on end-user’s privacy. Dangerous combinations of app permissions, and ad networks are used as features in our prediction models to understand the behavior of apps. Our models have a high classification accuracy of 95.9% considering the imbalance in real life between benign and malicious apps. Our assumption that certain app permissions can be a potential threat to the privacy of end users is confirmed to be one of the most impactful features of our prediction models. Since our study considers the impact of ad networks and malware permissions, it will help end-users make more informed decision about the app permissions they grant and understand that the app permissions open doors to more vulnerabilities, and at some point, benign apps can behave maliciously.

Keywords

• Authorization
• Application permission
• Security
• Privacy
• Android platform
• Human factors