Christine P. Chai's contribution to the Discussion of ‘Gaussian Differential Privacy’ by Dong et al.

By Jinshuo Dong, Aaron Roth, Weijie Su. 2021

206 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Sociology and Political Science

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Sociology and Political Science

Citations in the corpus

Tempered Sigmoid Activations for Deep Learning with Differential Privacy (2020, background): “Attempts to mitigate this have included new definitions of differential privacy (Dong, Roth, and Su 2019) or new learning algorithms (McMahan et al. 2018).”
Privacy in Deep Learning: A Survey (2020, positive): “s also employed by the works [102, 103, 104, 95, 96, 105] all of which use perturbed gradients. More recently, Bu et al. apply the Gaussian Differential Privacy (GDP) notion introduced by Dong et al. [121] to deep learning [122] to achieve a more reﬁned analysis of neural network training, compared ”
Differential Privacy for Deep and Federated Learning: A Survey (2022, background): “For the case of the Gaussian mechanism, the authors in [54] proved that: Corollary 2: The composition of k queries, each satisﬁes µ -GDP, is ( √ k µ ) -GDP.”
Secure, privacy-preserving and federated machine learning in medical imaging (2020, background): “Implementations range from simple random shuffling of the input data 57 to the introduction of noise to the dataset (Gaussian DP 58 with the benefit of better interpretability).”
Adversary Instantiation: Lower Bounds for Differentially Private Machine Learning (2021, background): “Since then, many works [13, 2, 54, 41, 31, 3] improved the analysis of the Moments Accountant analysis to get better theoretical privacy bounds.”
Differentially-private Federated Neural Architecture Search (2020, methodology): “With these in place, we present the following definitions from (Dong et al., 2019) for our proof.”
Enhanced Membership Inference Attacks against Machine Learning Models (2021, background): “, 2015] and has even motivated a new relaxed notion of differential privacy called f-DP [Dong et al., 2019].”
Large Language Models Can Be Strong Differentially Private Learners (2021, methodology): “We fine-tune these models with DP-Adam (Abadi et al., 2016; Kingma & Ba, 2014) (see Appendix A for details) and track privacy loss through Rényi DP (Mironov, 2017), but also report the converted from a Gaussian DP CLT (Dong et al., 2019) and from accurately composing tradeoff functions via fast Fou”