Table 6.
Current state-of-the-art open set RFF identification methods
Methods | Reference | Openness | Device/Protocol | Signal preprocessinga | Category of RFF/Further processing | Classifier | Experimental result |
---|---|---|---|---|---|---|---|
Traditional machine learning based | IEEE IoT 2021 [49] | 29.29% | WiMAX | Raw | RF-DNA/Relief-F for FDR | SVM | TPR ≥ 90%, FPR ≤ 10% at 6dB SNR |
IEEE IoT 2021 [97] | 18.35% | ZigBee | Frequency/ phase offset compensation | I/Q data/LDA for FDR | GPLDA | EER = 0.0063 at 30dB SNR | |
Deep learning based | IEEE TIFS 2021 [23] | 4.65% | ZigBee | Neural synchronization | I/Q data/Hypersphere representation | Auxiliary linear classifier | EER = 0.020, AUC = 0.998 at 30dB SNR with device agingb |
Generative model-based | IEEE GLOBECOM 2021 [120] | 50.00% | WiFi | Raw | I/Q data/Blind outlier generation with Autoencoder | OvA classifier and |A|+1 classification networkc | Testing accuracy: ≥84% |
(a)Specifically refers to preprocessing other than basic operations such as normalization and signal slicing, which do not require prior information. (b)Also known as parameter drift, which will be discussed in Section 8.1.1. (c)|A| denotes the number of known devices.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.