MAG-PUF: Magnetic Physical Unclonable Functions for Device Authentication in the IoT

Authenticating Internet of Things (IoT) devices is still a challenge, especially in deployments involving low-cost constrained nodes. The cited class of IoT devices hardly support dynamic re-keying solutions, hence being vulnerable to several attacks. To provide a viable general-purpose solution, in this paper we propose MAG-PUF, a novel lightweight authentication scheme based on the usage of unintentional magnetic emissions generated by IoT devices as Physical Unclonable Functions (PUFs). Specifically, through MAG-PUF, we collect unintentional magnetic emissions produced by the IoT devices at run-time while executing pre-defined reference functions, and we verify the match of such emissions with the profiles collected at enrolment time, providing device authentication. MAG-PUF enjoys unique flexibility, allowing the selection of an unlimited number and types of reference functions. We extensively assessed the performance of MAG-PUF through experiments on 25 Arduino devices and a set of exemplary reference functions. We obtained an authentication accuracy above 99%, hence proving the feasibility of using code-driven magnetic emissions as a lightweight, efficient, and robust PUF for IoT devices.