Device authentication over the wireless channel is still an open issue. This is especially true for low-end devices like the IoT ones, where the overhead required by traditional asymmetric cryptographic techniques can be overwhelming, or-more in general-when the crypto material might have been compromised. A robust solution for the above scenarios is Physical-Layer Authentication (PLA), which exploits the inherent intrinsic unique features of the wireless devices to achieve low-cost, crypto-less authentication. In this paper, we present Mag-Auth, a novel and lightweight authentication scheme that leverages the Electro-Magnetic (EM) emissions released at the joint connection between the wireless device and its antenna in response to an excitation signal. Specifically, Mag-Auth trains, on the collected EM emissions, an autoencoder and a Neural Network (NN). The autoencoder is employed to reject wireless devices that do not belong to the set the autoencoder and the NN have been trained over, while the NN is applied to uniquely identify the different classes of wireless transmitter-receiver pairs. Mag-Auth enjoys some unique features: it is privacy-preserving as it does not require to have access to the radio board (unlike, for instance, in-phase/quadrature (IQ)-based PLA methods); it caters to both wireless transmitter and receiver authentication scenarios; and, it sports striking performance. Indeed, our extensive experimental campaign involving 600 combinations of various wireless devices and antennas (including SDRs and IoTs) unveiled a minimum average F1-Score of 0.94 when classifying samples collected over a maximum length of 1s, proving the effectiveness and viability of using EM emissions as a lightweight, efficient, and robust authentication mechanism. Finally, we also released the collected EM emissions raw data to foster further investigations and development by Academia, Industry, and practitioners.
|Original language||English (US)|
|Title of host publication||WiSec 2023 - Proceedings of the 16th ACM Conference on Security and Privacy in Wireless and Mobile Networks|
|Publisher||Association for Computing Machinery, Inc|
|Number of pages||12|
|State||Published - May 29 2023|