Securing User Privacy in Cloud-Based Whiteboard Services Against Health Attribute Inference Attacks

Cloud-based whiteboard services have gained immense popularity, facilitating seamless collaboration and communication. However, the open-ended and persistent nature of whiteboard data exposes privacy vulnerabilities. In this paper, we investigate potential health attribute inference attacks that leverage drawings to infer sensitive user information without consent. We develop a local differentially private algorithm that perturbs drawings by spatially deforming them, providing provable privacy guarantees. Our algorithm is implemented in a software tool called DP-WhiteBoard. Extensive experiments demonstrate the algorithm’s ability to significantly reduce the accuracy of health attribute inference attacks while maintaining utility for benign recognition tasks. This work represents the first comprehensive study of emerging privacy threats in cloud-based whiteboards, proposing an scalable and adaptable solution with provable privacy guarantee. A demonstration of our work can be found at
https://youtu.be/5gD1Te1Fgnw
.