Papers › Backpropagation Clipping for Deep Learning with Differential Privacy

Backpropagation Clipping for Deep Learning with Differential Privacy

10 Feb 2022arXiv:2202.05089archive 2025-07-28

Timothy Stevens, Ivoline C. Ngong, David Darais, Calvin Hirsch, David Slater, Joseph P. Near

We present backpropagation clipping, a novel variant of differentially private stochastic gradient descent (DP-SGD) for privacy-preserving deep learning. Our approach clips each trainable layer's inputs (during the forward pass) and its upstream gradients (during the backward pass) to ensure bounded global sensitivity for the layer's gradient; this combination replaces the gradient clipping step in existing DP-SGD variants. Our approach is simple to implement in existing deep learning frameworks. The results of our empirical evaluation demonstrate that backpropagation clipping provides higher accuracy at lower values for the privacy parameter ϵ compared to previous work. We achieve 98.7% accuracy for MNIST with ϵ= 0.07 and 74% accuracy for CIFAR-10 with ϵ= 3.64.

PaperPDFCode

Code

uvm-plaid/backpropagation-clipping officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Deep LearningPrivacy PreservingPrivacy Preserving Deep LearningSensitivity

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

Gradient Clipping

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections