Bounding Training Data Reconstruction in DP-SGD
원문 초록
Differentially private training offers a protection which is usually interpreted as a guarantee against membership inference attacks. By proxy, this guarantee extends to other threats like reconstruction attacks attempting to extract complete training examples. Recent works provide evidence that if one does not need to protect against membership attacks but instead only wants to protect against training data reconstruction, then utility of private models can be improved because less noise is required to protect against these more ambitious attacks. We investigate this further in the context of DP-SGD, a standard algorithm for private deep learning, and provide an upper bound on the success of any reconstruction attack against DP-SGD together with an attack that empirically matches the predictions of our bound. Together, these two results open the door to fine-grained investigations on how to set the privacy parameters of DP-SGD in practice to protect against reconstruction attacks. Finally, we use our methods to demonstrate that different settings of the DP-SGD parameters leading to the same DP guarantees can result in significantly different success rates for reconstruction, indicating that the DP guarantee alone might not be a good proxy for controlling the protection against reconstruction attacks.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
TimberStrike: Dataset Reconstruction Attack Revealing Privacy Leakage in Federated Tree-Based Systems
Training Set Reconstruction from Differentially Private Forests: How Effective is DP?
Investigating Privacy Leakage in Dimensionality Reduction Methods via Reconstruction Attack
Face Reconstruction Transfer Attack as Out-of-Distribution Generalization
URVFL: Undetectable Data Reconstruction Attack on Vertical Federated Learning
Trained Random Forests Completely Reveal your Dataset
토론
이 논문과 연결된 포럼 글이 아직 없습니다.