Tianhao Wang

The paper proposes DP-GCL, a novel differentially private contrastive learning framework that improves representation learning on sensitive data by bounding gradient dependency through localized group-level negative sampling.

The paper proposes Multi-Recall Memory MIA (MRMMIA), a unified attack framework to test for privacy leakage by determining if a candidate memory unit belongs to a chat agent's private memory store.

DP-SAPF introduces a saliency-aware parameter fine-tuning method that selectively identifies the most critical parameters for LoRA training, significantly improving the utility and fidelity of differentially private image synthesis while reducing computational cost.