HVPPF: A Hierarchical Visual Privacy Protection Framework for Cloud Services Customized to Balance Privacy and Usability

With the growing demand for cloud services, traditional image privacy encryption methods applied in cloud scenarios reveal two major issues. First, security is often achieved at the expense of visibility, which is incompatible with cloud service scenarios such as information preview and search. Second, there is a lack of design for hierarchical visual privacy for users with different security levels. For the above problems, a multi-level key mechanism is designed and integrated with the YoloV5 network, providing not only multi-level privacy protection for sensitive regions but also achieving a balance between visibility and security in these regions. Simulation results demonstrate that the proposed framework can decrypt images with multi-level visual effects. Performance analysis shows that the framework achieves an adjustable balance between visibility and security, which users can modify by adjusting parameters. Compared to other visibility-security trade-off schemes, this approach offers advantages including strong reversibility, high image size compatibility, adjustable visual effects, and computational efficiency.