YOLOv8 for Personal Device Screen Detection to Preserve Data Privacy: Enhancing Security Measures in Real-Time Monitoring Systems
Apichaya Nimkoompai, Yuenyong Nilsiam, Siranee Nuchitprasitchai, Puwadol Sirikongtham · 2025
In public and semi-public environments, personal device usage often exposes sensitive information, presenting a challenge to user privacy when such data is captured by surveillance systems. This study investigates the application of YOLOv8, an advanced object detection model, to identify and obscure personal electronic device screens, particularly mobile phones, in CCTV footage. Utilizing real-time blurring with a Gaussian kernel, this approach ensures that sensitive information displayed on these devices remains unreadable in video footage. The study aims to address the following research questions: (1) Each version of YOLOv8 varies in its effectiveness for detecting and identifying mobile device screens? (2) What level of Gaussian blurring is sufficient to obscure sensitive information without compromising processing efficiency? Experimental results demonstrate that YOLOv8m consistently excels, particularly in precision and mAP50, making it the ideal choice for applications where high accuracy is essential. Meanwhile, YOLOv8n, as the smallest model, delivers promising outcomes in terms of processing speed. Additionally, a 35x35 Gaussian kernel achieves an optimal balance between privacy protection and computational efficiency. These findings support the feasibility of integrating YOLOv8-based real-time blurring into surveillance systems, providing a robust solution for privacy protection in high-risk surveillance areas.