SenSig: Practical IoT Sensor Fingerprinting Using Calibration Data

Abstract

Sensing technologies are becoming ever more ubiquitous in society and increasingly finding their way into important and intimate aspects of our lives such as Industrial Internet of Things (IIOT) and Smart Homes. Accordingly, it's vital to fingerprint these sensors and devices enabling the detection of any malicious or malfunctioning sensors that may be present. The aim of this paper is to provide a simple and lightweight means of fingerprinting motion sensors, and by extension the devices these sensors reside in. To generate our fingerprints, we use the data produced by motion sensors (more specifically, the gyroscope) during the calibration process on start-up. Subsequently, we employ the use of a novel form of quantisation, and various signal processing techniques on this sensor data to generate our sensor fingerprints. Our results show that such calibration data is fingerprintable, and we demonstrate the effectiveness of a potential use case of our fingerprints: identification, where we are able to uniquely identify a sensor with a 0 % EER and 38-bits of entropy.