An Multi-Dimensional Feature Fusion Deep Learning Based Wireless Device Identification

As security concerns in smart home wireless communication systems continue to grow, existing device identification technologies face significant challenges, including inadequate characterization of device features due to complex multipath propagation and high false alarm rates caused by low intra-class identification accuracy. In this paper, we propose a deep feature fusion RF fingerprint identification model that combines feature fusion with SE attention mechanism. By integrating 13 expert features with features extracted by 1DCNN and leveraging the SE attention mechanism for optimal feature weight, the proposed model enhances the recognition rate of smart home devices. Experimental results on public datasets demonstrate that the method achieves an identification accuracy of 99.84 %, effectively resolving the problem of insufficient device recognition in smart home environments.