Nearly Tight Black-Box Auditing of Differentially Private Machine Learning
This provides a practical tool for detecting privacy violations and improving DP-SGD analysis, though it is incremental as it builds on existing auditing methods.
The paper tackles the problem of auditing Differentially Private Stochastic Gradient Descent (DP-SGD) in a black-box threat model, achieving tighter empirical privacy estimates than prior work, with results such as ε_emp = 7.21 on MNIST and 6.95 on CIFAR-10 for theoretical ε=10.0.
This paper presents an auditing procedure for the Differentially Private Stochastic Gradient Descent (DP-SGD) algorithm in the black-box threat model that is substantially tighter than prior work. The main intuition is to craft worst-case initial model parameters, as DP-SGD's privacy analysis is agnostic to the choice of the initial model parameters. For models trained on MNIST and CIFAR-10 at theoretical $\varepsilon=10.0$, our auditing procedure yields empirical estimates of $\varepsilon_{emp} = 7.21$ and $6.95$, respectively, on a 1,000-record sample and $\varepsilon_{emp}= 6.48$ and $4.96$ on the full datasets. By contrast, previous audits were only (relatively) tight in stronger white-box models, where the adversary can access the model's inner parameters and insert arbitrary gradients. Overall, our auditing procedure can offer valuable insight into how the privacy analysis of DP-SGD could be improved and detect bugs and DP violations in real-world implementations. The source code needed to reproduce our experiments is available at https://github.com/spalabucr/bb-audit-dpsgd.