Progressive physical-layer fingerprinting cuts Ethernet auth data
Original title: Progressive-Observation Physical-Layer Fingerprinting for Efficient Open-Set Ethernet Authentication
On held-out days the method cuts waveform data 75.1% and Raspberry Pi 5 CPU time 73.7% without raising error rates.
The paper proposes progressive-observation physical-layer fingerprinting so edge nodes do not process a fixed-length waveform for every IoT authentication query. A small CNN rejects obvious unknowns from a short capture, spectral matching accepts clear registered devices from a slightly longer one, and only ambiguous queries use the full waveform. An adaptive calibration realigns features and thresholds from a few same-day references so the CNN is not retrained when fingerprints drift. On 100BASE-TX devices of two controller types, false rejection and false acceptance stay equal or lower than fixed-length processing, while an edge node can serve about four times as many queries at the same monitoring bandwidth.
- Waveform data per query falls 75.1% versus fixed-length processing.
- CPU time on a Raspberry Pi 5 falls 73.7% on average.
- Calibration uses current-day references from registered devices instead of retraining the CNN.
From the original article
1 min read at arxiv.orgPhysical-layer fingerprinting (PLF) has emerged as a promising authentication approach for Internet of Things devices. However, most PLF systems acquire and process a fixed-length waveform for every authentication query, burdening resource-constrained edge nodes. To address this issue, we propose a progressive-observation PLF method that adapts the waveform length to each query.
By Chenming Zhang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.