Multiscale FFT algorithm improves privacy loss accounting for random allocation
Multiscale FFT reduces runtime from quadratic to O(α^{-1} log(1/α)) and rounding errors to O(α^2/log(t)) for DP-SGD with PREAMBLE.
Stories per day · 30 days
4 RSS
Relateddifferential-privacy5dp-sgd2celeba1cifar-101diffusion-models1fft1integer-releases1llm1noise-schedule1optimization1privacy1rag1random-allocation1randomized-rounding1survey1
8 stories, newest first · one card per story
Stories tagged #privacy-accounting
Paper derives optimal privacy-budget schedules for private SGD
PB-AMD uses adjoint gradients and entropic updates so noise schedules stay positive and spend the certified zCDP budget.
Paper accounts privacy loss during diffusion-model sampling
On CIFAR-10 and CelebA-64, loss rises sharply while semantic content forms, motivating hybrid denoisers.
Paper states exact resource laws for hidden integer feedback
Keeping carries until an observed echo yields exact compilation with support-width exponent 3C-2 for budget C at least 2.
From One Query to Many: A Critical Survey of Privacy Accounting in RAG
Survey says RAG differential-privacy claims cannot be compared by epsilon alone across repeated queries.
Better Batches, Better Privacy: Random Allocation for DP-BandMF
Random-allocation batching for DP-BandMF cuts RMSE up to 1.85 times and speeds privacy accounting on GPUs.
Closed-Form Noise Calibration Against Membership Inference for Random-Allocation DP-SGD
A closed-form noise formula bounds membership-inference accuracy for random-allocation DP-SGD and needs less noise.
Differential Privacy of Gradient Descent on Perturbed Objectives
Proves explicit privacy bounds for finite gradient-descent iterates on perturbed objectives, removing ambient-dimension factors for generalized linear models.