Class-Incremental Learning for Wireless Device Identification in IoT

Deep learning (DL) has been utilized pervasively in the Internet of Things (IoT). One typical application DL IoT is device identification from wireless signals, namely, noncryptographic (NDI). However, components NDI systems have to evolve adapt operational variations, such a paradigm termed as incremental (IL). Various IL algorithms proposed and many them require dedicated space store increasing amount historical data, therefore, they are not suitable for or mobile applications. Besides, conventional schemes can provide satisfying performance when data available. In this article, we address problem new perspective, first, metric measure degree topological maturity DNN models conflict class-specific fingerprints. We discover that an important cause degradation IL-enabled owing devices’ Second, also show lead low systems. Thirdly, propose channel separation-enabled (CSIL) scheme without using which our strategy automatically separate fingerprints different stages avoid potential conflict. Finally, evaluated effectiveness framework real automatic-dependent surveillance-broadcast (ADS-B), aviation. The be applied accurate devices variety applications services. Data code available at IEEE Dataport (DOI: 10.21227/1bxc-ke87) https://github.com/pcwhy/CSIL .