Fingerprinting smartphones from embedded transducers

Abstract

Due to the significant privacy risks that smartphones present and due to their importance in device authentication and forensics investigations, fingerprinting smartphones have become increasingly popular. This thesis is focused on accelerometers, loudspeakers, microphones and camera sensors as potential fingerprint sources for smartphone embedded transducers. While there is little user knowledge regarding the privacy dangers, the output of these transducers, which convert one form of energy into another, leaks across numerous channels, like social networks, mobile apps and cloud services. Several signal processing techniques are used to extract characteristics and various traditional machine-learning algorithms are employed to fingerprint different and identical sensors. This thesis also proposes a system for device pairing based on accelerometer data collected from several transportation modes, i.e., tram, train, car, bike, walk and shake. In addition to smartphone sensors fingerprinting, ECU (Electronic Control Unit) fingerprinting is discussed as an extension.