Preamble-based RF-DNA Fingerprinting Under Varying Temperatures

A total of 30.9 billion Internet of Things (loT) deployments are expected by 2025 with most employing weak or no encryption at all, which raises concerns about loT security. This concern is exacerbated by loT-connected critical infrastructure and the successful exploitation of this security vulnerability. This led researchers to propose a physical layer-based loT security solution coined Specific Emitter Identification (SEI). However, SEI has been shown to be sensitive to temperature changes. This sensitivity is important when considering loT deployments in highly variable temperature environments. The presented approach shows the temperature sensitivity of SEI is mitigated when the classifier is trained using RF-DNA fingerprints drawn from waveforms collected at two temperatures. In fact, SEI performance improves the most when the two temperatures are at or near the extremes of the operating temperature range. Specifically, our work shows that training SEI classifiers using the extremes of the collected temperatures improves overall classification performance across temperature ranges. The work in this paper also shows that emitters operating in a sub-ambient, exothermic state have a more consistent fingerprint than those operating in a high-temperature, endothermic state.