Droid Mood Swing (DMS): Automatic Security Modes based on Contexts

Abstract

Smartphones are becoming ubiquitous and we use them for different types of tasks. One problem of using the same device for multiple tasks is that each task requires a different security model. To address this problem, we introduce Droid Mood Swing (DMS), an operating system component that applies different security policies to detected security modes automatically. DMS uses a context manager that tracks the context of the phone from the available sensors. DMS then determines the security mode from the contexts and can impose a number of security measures, namely fine-grained permissions, an intent firewall, a context-aware SD card filesystem, and a permission verification system. The permission verification system uses machine learning techniques to detect suspicious apps and anomalous permission requests. DMS also provides an API that enables third-party developers to make their apps behave differently in different modes. DMS is designed especially for end users and does not compromise the usability of the phone. Device vendors will be able to control configurations (a switching logic and security policies) of the modes through DMS. We implement DMS using the Android Open Source Project (AOSP) and evaluate it in terms of portability, functionality, security, and operational overheads. The evaluation results show that DMS offers a more secure smartphone operating system without incurring any noticeable overhead.