Multi-scale Attentional Temporal Memory Fusion Network For Radio Frequency Fingerprint Identification Under Low SNR

Document Type

Article

Publication Date

9-5-2024

Identifier/URL

41552852 (Pure)

Find this in a Library

Catalog Record

Abstract

With the explosive growth of the number of internet of things (IoT) devices, the identity authentication of IoT devices has become a hot research direction of IoT communication security. Radio frequency fingerprint (RFF) identification is an effective solution to the identity authentication of IoT devices, but its performance declines significantly at low signal-to-noise ratio (SNR). In this paper, a parallel multi-scale attentional temporal memory fusion network (MA-TMFN) is designed from the perspectives of feature extraction and feature fusion. The network extracts two scale features, global features and local features, and introduces an attentional feature fusion (AFF) module to assign different weights to the global and local features, so that the network can focus important multi-scale features of radio frequency (RF) signals in low SNR scenarios, thereby improving the RFF identification rate. Experiments were conducted with 18 IoT devices and the public automatic dependent surveillance-broadcast (ADS-B) dataset to evaluate the performance of MA-TMFN. Experimental results show that the proposed MA-TMFN performs better than other existing deep learning networks on different datasets at low SNR.

DOI

10.1109/ICCCAS62034.2024.10652830

Link to Full Text

Catalog Record

Share

COinS