Using Audio Characteristics for Mobile Device Authentication

This work attempts to use the innate manufacturing defects in hardware components as identification characteristics for mobile phones. Different components of mobile phones related to I/O operations, such as sensors, were assessed for suitability. From this process, efforts were focused on using both the phone’s speaker and microphone in combination to generate samples containing hardware defects which could then be classified. In our approach, a known audio was played using a cellphone’s speakers and recorded using the same device’s speaker, creating an audio sample. Multiple different groups of samples were taken to test the impact of certain variables on the sample accuracy. The collected samples then had their frequency responses extracted and classified. Different classifiers were used to classify samples with some configurations of classifiers and sample groups achieving over 99.9% accuracy. The results presented in this paper indicate that the manufacturing defects in speakers and microphones could potentially be used for the purposes of device identification.