Identifying Bluetooth Low Energy Devices via Physical-Layer Hardware Impairments

Abstract: It is desirable to be able to distinguish between Bluetooth Low Energy (BLE) devices purely based on individual hardware imperfections because such imperfections are often prohibitly expensive to imitate. Other works have been done on this topic in the past, which utilized signal transients requiring a multi-GHz sampling rate recorded with costly special equipment to be satisfactorily captured. This work explores an alternative approach which avoids these requirements, with new features tailor-made for the Bluetooth Low Energy protocol that allow us to achieve a micro-recall of 99% with the full test set of 58 motes. The lessons learned during this work bring us one step closer to pervasive deployment where regular BLE-enabled hardware could be configured to use radiometric fingerprinting in order to distinguish between devices in receiver range.