Privacy in Collaborative ML

Sep 27, 2018

Differential Privacy

Machine Learning Differential Privacy Selected

Junyuan Hong

Postdoctoral Fellow

My research interest lies in the interaction of human-centered AI and healthcare.

Publications

Shake to Leak: Fine-tuning Diffusion Models Can Amplify the Generative Privacy Risk

We propose a new risk to published generative models that finetuning on generated samples can exacerbate the privacy leakage.

Zhangheng Li, Junyuan Hong, Bo Li, Zhangyang Wang

Shake to Leak: Fine-tuning Diffusion Models Can Amplify the Generative Privacy Risk

DP-OPT: Make Large Language Model Your Privacy-Preserving Prompt Engineer

We make local LLMs to engineer privacy-preserving prompts that are transferrable for cloud models.

Junyuan Hong, Jiachen T. Wang, Chenhui Zhang, Zhangheng Li, Bo Li, Zhangyang Wang

DP-OPT: Make Large Language Model Your Privacy-Preserving Prompt Engineer

Understanding Deep Gradient Leakage via Inversion Influence Functions

We propose a new metric to efficiently evaluate the privacy risks from gradient inversion and provides new insights.

Haobo Zhang, Junyuan Hong, Yuyang Deng, Mehrdad Mahdavi, Jiayu Zhou

Understanding Deep Gradient Leakage via Inversion Influence Functions

Outsourcing Training without Uploading Data via Efficient Collaborative Open-Source Sampling

We propose a new privacy-preserving learning framework, outsourcing training to cloud without uploading data, which provides more data without injecting noise into gradient or samples.

Junyuan Hong, Lingjuan Lyu, Jiayu Zhou, Michael Spranger

Outsourcing Training without Uploading Data via Efficient Collaborative Open-Source Sampling

Dynamic Privacy Budget Allocation Improves Data Efficiency of Differentially Private Gradient Descent

Protecting privacy in learning while maintaining the model performance has become increasingly critical in many applications that …

Junyuan Hong, Zhangyang Wang, Jiayu Zhou

Learning Model-Based Privacy Protection under Budget Constraints

Protecting privacy in gradient-based learning has become increasingly critical as more sensitive information is being used. Many …

Junyuan Hong, Haotao Wang, Zhangyang Wang, Jiayu Zhou