Real-Time Privacy Preservation for Robot Visual Perception

2025-05-08

AI Summary

Real-time robotic video streams risk privacy leakage via sensitive objects (e.g., faces), yet existing methods struggle to simultaneously ensure complete occlusion and real-time performance. Method: We propose PCVS—a real-time privacy-preserving framework integrating formal logical specifications with online visual detection. Its core innovation is the first extension of conformal prediction theory to multi-frame video sequences, enabling adaptive, verifiable probabilistic lower bounds for logic-based privacy specifications (e.g., “no face shall ever appear in the frame”). PCVS combines YOLO-based detection with dynamic blurring to achieve low-latency processing without compromising downstream task performance. Results: Experiments across multiple datasets show PCVS achieves >95% specification satisfaction rate—consistently exceeding its theoretical lower bound—and has been successfully deployed on a physical robotic platform.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Privacy
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Many robots (e.g., iRobot's Roomba) operate based on visual observations from live video streams, and such observations may inadvertently include privacy-sensitive objects, such as personal identifiers. Existing approaches for preserving privacy rely on deep learning models, differential privacy, or cryptography. They lack guarantees for the complete concealment of all sensitive objects. Guaranteeing concealment requires post-processing techniques and thus is inadequate for real-time video streams. We develop a method for privacy-constrained video streaming, PCVS, that conceals sensitive objects within real-time video streams. PCVS takes a logical specification constraining the existence of privacy-sensitive objects, e.g., never show faces when a person exists. It uses a detection model to evaluate the existence of these objects in each incoming frame. Then, it blurs out a subset of objects such that the existence of the remaining objects satisfies the specification. We then propose a conformal prediction approach to (i) establish a theoretical lower bound on the probability of the existence of these objects in a sequence of frames satisfying the specification and (ii) update the bound with the arrival of each subsequent frame. Quantitative evaluations show that PCVS achieves over 95 percent specification satisfaction rate in multiple datasets, significantly outperforming other methods. The satisfaction rate is consistently above the theoretical bounds across all datasets, indicating that the established bounds hold. Additionally, we deploy PCVS on robots in real-time operation and show that the robots operate normally without being compromised when PCVS conceals objects.

Problem

• Concealing privacy-sensitive objects in real-time robot video streams
• Ensuring complete concealment with logical specifications and detection models
• Providing theoretical bounds for specification satisfaction via conformal prediction

Innovation

• PCVS method conceals sensitive objects in real-time video
• Uses logical specifications to constrain privacy-sensitive objects
• Conformal prediction ensures theoretical bounds on object concealment