Sensor Identification via Acoustic Physically Unclonable Function

Abstract

The traceability of components on a supply chain from a production facility to deployment and maintenance depends upon its irrefutable identity. There are two well-known identification methods: an identity code stored in the memory and embedding custom identification hardware. While storing the identity code is susceptible to malicious and unintentional attacks, the approach of embedding a custom identification hardware is infeasible for sensor nodes assembled with Commercially-Off-the-Shelf devices. We propose a novel identifier - Acoustic PUF based on the innate properties of the sensor node. Acoustic PUF combines the uniqueness component and the position component of the sensor device signature. The uniqueness component is derived by exploiting the manufacturing tolerances, thus making the signature unclonable. The position component is derived through acoustic fingerprinting, thus giving a sticky identity to the sensor device. We evaluate Acoustic PUF for Uniqueness, Repeatability, and Position identity with a deployment spanning several weeks. Through our experimental evaluation and further numerical analysis, we prove that Acoustic PUF can uniquely identify thousands of devices with 99% accuracy while simultaneously detecting the change in position. We use the physical position of a device within a synthetic sound-field both as an identity measure as well as to validate physical integrity of the device.