In This Section

Project Overview    

Modern devices have become “smart” in recent years with the advancement of modern electronics and wireless communication systems. Smart devices such as smartphone, smartwatch, fitness trackers, etc. are equipped with high precision sensors, empowering them to gather information about user characteristics as well as the surrounding environment. These sensor-enriched devices have opened a new domain of sensing-enabled applications. With the rapid growth sensing-enabled applications, smart devices are integrated in every possible application domain, from home security to health care to military. Some applications can even learn characteristics of users using sensor data and take automatic decisions to improve the user experience.

Nevertheless, integration of different sensors in smart devices has introduced a novel way to exploit these devices. Using sensors an adversary can successfully attack a smart device by (1) triggering existing malware, (2) transferring malware, (3) leaking sensitive information, or (4) stealing valuable information. These sensor-based threats expose flaws in existing sensor management systems of smart devices. In this project, we investigate and develop a context-aware IDS, 6thSense, to address these sensor-based threats. Our proposed framework is built upon the observation that for any user activity or task in smart devices, a different, but a specific set of sensors becomes active. 6thSense monitor sensor data in real-time and learn how sensors’ states change with different activities. In training phase, we collect sensor data for different user activity and build a context-aware model of a task which is the ground truth of the framework. In the detection phase, 6thSense observes sensor data and checks against the ground truth. In 6thSense, three different approaches are used for detection method: Markov Chain, Naïve Bayes Model, and standard Machine Learning (ML) Techniques. Our proposed framework provides real-time sensor monitoring and shows higher accuracy in detecting sensor-based threats.