Individual privacy accounting for differentially private stochastic gradient descent
Differentially private stochastic gradient descent (DP-SGD) is the workhorse algorithm for
recent advances in private deep learning. It provides a single privacy guarantee to all …
recent advances in private deep learning. It provides a single privacy guarantee to all …
Conan: Distributed Proofs of Compliance for Anonymous Data Collection
We consider how to design an anonymous data collection protocol that enforces compliance
rules. Imagine that each client contributes multiple data items (eg, votes, location crumbs, or …
rules. Imagine that each client contributes multiple data items (eg, votes, location crumbs, or …
Beyond Statistical Estimation: Differentially Private Individual Computation in the Shuffle Model
The shuffle model of differential privacy (DP) has recently emerged as a powerful one for
decentralized computation without fully trustable parties. Since it anonymizes and permutes …
decentralized computation without fully trustable parties. Since it anonymizes and permutes …
Differentially Private Numerical Vector Analyses in the Local and Shuffle Model
S Wang, S Yu, X Ren, J Li, Y Li… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
Numerical vector aggregation plays a crucial role in privacy-sensitive applications, such as
distributed gradient estimation in federated learning and statistical analysis of key-value …
distributed gradient estimation in federated learning and statistical analysis of key-value …
Personalized Differential Privacy in the Shuffle Model
R Yang, H Yang, J Fan, C Dong, Y Pang… - … Conference on Artificial …, 2023 - Springer
Personalized local differential privacy is a privacy protection mechanism that aims to
safeguard the privacy of data by using personalized approaches, while also providing …
safeguard the privacy of data by using personalized approaches, while also providing …