Your Robotic Desktop Companion: Privacy-Preserving Webcam for Video Conferencing via Motion-Augmented Human Perception

Video conferencing has become an integral part of everyday communication. However, users often lose awareness of their active webcams during online meetings, resulting in inadvertent privacy leakage. Existing webcam indicators, such as LED lights and software icons, demand constant user vigilance, which becomes impractical and ineffective during long online meetings. Our survey with 85 participants confirms the insufficiency of current webcam privacy mechanisms. To address this security gap, we propose a robotic desktop assistant that communicates camera activity through subtle physical motion, turning toward the user to mimic natural eye contact. The core concept is to leverage humans’ innate sensitivity to environmental motion for intuitive human-camera interaction, eliminating the need for users to consciously look at the camera to check its status. We implement the robotic desktop assistant concept based on Raspberry Pi and a servo-mounted webcam. When the webcam is on, the system operates a closed-loop controller that tracks user motion using a lightweight computer vision model and adjusts its pan-tilt angles in real time to mimic human-like eye contact. The loop latency is less than 45ms. The system can be deployed both as a peripheral webcam or an independent assistant. Experimental results and user studies show that the motion-based cues significantly improve users’ awareness of webcam activity without distraction or adding to user workload. The prototype system can be further enhanced by integrating voice control, large language models, and a laser projection module, transforming it into a fully functional, standalone online meeting system, like your intelligent desktop companion.