SmartDog: Real-time Detection of Smartphone Theft

With the enhanced computing capabilities and a wide variety of functions available on smartphones, critical and sensitive information, such as contact lists, messages, schedules, credit card numbers, is stored on smartphones, which makes preventing smartphone from being stolen of an unprecedented importance. Loss of smartphones not only causes economic loss but also jeopardizes the privacy of the owners. Existing anti-theft schemes only provide passive protection, which remotely manipulates lost phones to lock and delete private information. However, if the stealer deletes the protection apps or shuts down the phone before the owner being aware that the phone has been stolen, remedial actions become invalid. Hence it is of great importance to detect smartphone theft as soon as it happens, such that stealing can be stopped. In this paper, we propose SmartDog, a real-time smartphone anti-theft scheme for keeping smartphone safe. With embedded motion sensors, SmartDog can effectively capture the unique and reliable biometrical features of the owners about how they pick up the smartphone from their pockets or bags. If a stealer tries to steal a smartphone away from the pocket or bag of the owner, SmartDog will detect the unusual motion, even if an attacker can see an owner picking up his/her phone (Since the attacker can hardly reproduce the same behavior). We implement SmartDog and conduct intensive trace-driven simulations with 'picking up' samples from 20 volunteers, collected over two weeks. SmartDog achieves 10.2% average false positive and 5.5% average false negative error rates, respectively.