Incremental Learning for Radio Frequency Fingerprint Identification

With the rapid development of Internet of Things technology, wireless communication become an essential part in every field, which also bring about many wireless communication security problems. Traditional solutions to wireless communication security problems are mostly at the software level and protocol level, ignoring the physical characteristics of the device itself. Radio frequency fingerprint (RFF) can distinguish different devices in the physical level. Most of the existing incremental learning based radio frequency fingerprint identification (RFFI) are need a large amount of old data. In this paper, we review lots of RFFI method based on ML, DL or IL, and summarize a generic framework for RFFI, and propose our method to efficiently reduce the needed amount of old data in IL based RFFI, which saves training time and storage space.