A Privacy Enhancing Technique to Evade Detection by Street Video Cameras Without Using Adversarial Accessories

2025-01-26

AI Summary

This study addresses the challenge of wearable-free privacy protection in street-level video surveillance, identifying and quantifying location- and illumination-dependent blind spots in mainstream pedestrian detectors (e.g., Faster R-CNN). Method: We propose a geometric–photometric joint modeling framework to precisely localize low-confidence detection regions, and design a minimum-confidence path planning algorithm to guide pedestrians through detection-weak zones. Additionally, we introduce a countermeasure enhancement strategy involving confidence calibration and detector refinement to improve robustness. Contribution/Results: Experiments show that the planned paths reduce maximum and mean detection confidence by 0.09 and 0.13, respectively; after countermeasures, true positive rate and mean confidence increase by 0.03 and 0.15. This work is the first to systematically characterize the spatial sensitivity of pedestrian detectors, establishing a novel paradigm for environment-aware, passive privacy protection.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Privacy
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

In this paper, we propose a privacy-enhancing technique leveraging an inherent property of automatic pedestrian detection algorithms, namely, that the training of deep neural network (DNN) based methods is generally performed using curated datasets and laboratory settings, while the operational areas of these methods are dynamic real-world environments. In particular, we leverage a novel side effect of this gap between the laboratory and the real world: location-based weakness in pedestrian detection. We demonstrate that the position (distance, angle, height) of a person, and ambient light level, directly impact the confidence of a pedestrian detector when detecting the person. We then demonstrate that this phenomenon is present in pedestrian detectors observing a stationary scene of pedestrian traffic, with blind spot areas of weak detection of pedestrians with low confidence. We show how privacy-concerned pedestrians can leverage these blind spots to evade detection by constructing a minimum confidence path between two points in a scene, reducing the maximum confidence and average confidence of the path by up to 0.09 and 0.13, respectively, over direct and random paths through the scene. To counter this phenomenon, and force the use of more costly and sophisticated methods to leverage this vulnerability, we propose a novel countermeasure to improve the confidence of pedestrian detectors in blind spots, raising the max/average confidence of paths generated by our technique by 0.09 and 0.05, respectively. In addition, we demonstrate that our countermeasure improves a Faster R-CNN-based pedestrian detector's TPR and average true positive confidence by 0.03 and 0.15, respectively.

Problem

• Privacy Protection
• Pedestrian Detection
• Surveillance Evasion

Innovation

• Privacy Protection
• Pedestrian Detection Accuracy
• Adversarial Robustness