At this years ICNP 2020 IEEE conference a paper titled "Real-World ADS-B signal recognition based on Radio Frequency Fingerprinting" (pdf file) was presented by researchers from the University of Harbin, China. The idea presented in the paper is to use RF "fingerprinting" techniques to uniquely identify and confirm that the ADS-B signal originates from the correct aircraft source.
RF fingerprinting works on the premise that every transmitter has small manufacturing variances that result in slightly different signals be transmitted, resulting in a unique "fingerprint" that can be traced to a particular transmitter. The idea here is to use these fingerprints to ensure that a known aircraft is indeed transmitting an ADS-B signal and the signal is not being transmitted from a fake spoofer. ADS-B is completely unencrypted and not authenticated, so spoofing of ADS-B signals may be a real security threat.
In the teams research they use an RTL-SDR to collect ADS-B signals from five different aircraft. They then use that data to create "Contour Stellar Images" and train a deep learning neural network which after training accurately identifies which aircraft a signal comes from.
In previous posts we've seen the idea of fingerprinting used by Disney research and others to identify electronic devices, to authenticate RF IoT devices and to identify handheld transmitters via CTCSS fingerprints.