RateCount: Learning-Free Device Counting by Wi-Fi Probe Listening

2025-07-04

AI Summary

Traditional machine learning-based device counting methods suffer from low deployment efficiency under Wi-Fi MAC address randomization, requiring extensive data cleaning, model training, and hyperparameter tuning. Method: This paper proposes a training-free, lightweight, and theoretically unbiased device counting method. It leverages the probe request frame arrival rate observed by an access point (AP) within a time window to derive a closed-form estimator. Rigorous error modeling and variance lower-bound analysis ensure statistical reliability, while a device-to-person calibration mechanism enables natural scalability to people counting. Results: Evaluated across diverse real-world scenarios, the method achieves device counting accuracy comparable to state-of-the-art (SOTA) learning-based approaches, while significantly outperforming them in person counting. Crucially, it incurs zero training overhead, thereby overcoming a key bottleneck for efficient deployment under privacy-preserving constraints.

Technology Category

• Machine Learning: Privacy
• Humans and AI: Interaction Techniques and Devices
• Intelligent Robots: State Estimation

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments

Abstract

A Wi-Fi-enabled device, or simply Wi-Fi device, sporadically broadcasts probe request frames (PRFs) to discover nearby access points (APs), whether connected to an AP or not. To protect user privacy, unconnected devices often randomize their MAC addresses in the PRFs, known as MAC address randomization. While prior works have achieved accurate device counting under MAC address randomization, they typically rely on machine learning, resulting in inefficient deployment due to the time-consuming processes of data cleaning, model training, and hyperparameter tuning. To enhance deployment efficiency, we propose RateCount, an accurate, lightweight, and learning-free counting approach based on the rate at which APs receive PRFs within a window. RateCount employs a provably unbiased closed-form expression to estimate the device count time-averaged over the window and an error model to compute the lower bound of the estimation variance. We also demonstrate how to extend RateCount to people counting by incorporating a device-to-person calibration scheme. Through extensive real-world experiments conducted at multiple sites spanning a wide range of counts, we show that RateCount, without any deployment costs for machine learning, achieves comparable counting accuracy with the state-of-the-art learning-based device counting and improves previous people counting schemes by a large margin.

Problem

• Counts Wi-Fi devices without MAC address randomization issues
• Eliminates need for machine learning in device counting
• Extends to people counting with device-to-person calibration

Innovation

• Learning-free Wi-Fi device counting via PRF rate
• Unbiased closed-form expression for device estimation
• Device-to-person calibration for people counting