Characterizing Wi-Fi Probing Behavior for Privacy-Preserving Crowdsensing

Abstract

Smartphones and the signaling messages they emit allow third parties to learn about the owners' mobility. While Wi-Fi and Bluetooth signaling messages have been (mis)used for tracking individuals, there are also privacy-respecting uses: crowd sensing for estimating the number of people in an area and their dynamics, is one such example. However, the very useful countermeasures against individual tracking, most prominently MAC address randomization, also complicate crowd size estimation. In this paper, we present an online estimation algorithm that operates only on ephemeral MAC addresses and, if desired, signal strength information to distinguish relevant signals from background noise. We use measurements and simulations to calibrate our counting algorithm and collect numerous data sets which we use to explore the algorithm's performance in different scenarios.