Privacy-Preserving Randomized-MAC WiFi Client Counting with Short-Term-Coherent Waveform Features and a Bayesian Information Criterion

The randomization of WiFi MAC addresses mandated by recent data privacy legislation has created an urgent need for audience estimation techniques that can function without unique MAC addresses. Proposed solutions based on the statistics of Probe Requests require expensive auxiliary calibration systems, while attempts to de-randomize MAC addresses can be easily defeated with software patches, and are considered by many to run counter to privacy directives. The present work proposes a new solution, based on short-term coherence of Probe Request waveforms and a Bayesian Information Criterion, that is inherently privacy-respecting and can be implemented in a standard WiFi access point. Results incorporating field measurements into a realistic simulation indicate a promising new approach to privacy-preserving WiFi client counting.