Efficient DP-SGD for LLMs with Randomized Clipping

• URL: http://arxiv.org/abs/2605.24879v1
• Date: Sun, 24 May 2026 05:44:20 GMT
• Title: Efficient DP-SGD for LLMs with Randomized Clipping
• Abstract summary: Large language models (LLMs) are trained on vast datasets that may contain sensitive information.<n>We propose DP-SGD-RC, a novel variant of DP-SGD with randomized clipping that reduces memory and compute complexity.<n>We provide a tight privacy analysis showing that DP-SGD-RC achieves noise competitive with deterministic clipping.
• Score: 28.00696116795116
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) are trained on vast datasets that may contain sensitive information. Differential privacy (DP), the de facto standard for formal privacy guarantees, provides a principled framework for training LLMs with provable privacy protection. However, state-of-the-art DP training implementations rely on fast gradient clipping techniques with memory overhead $O(B \min\{T^2, d^2\})$, where $B$ is the batch size, $T$ is the sequence length, and $d$ is the model width. This becomes prohibitive as both model size and context length grow. We propose DP-SGD-RC, a novel variant of DP-SGD with randomized clipping that reduces memory and compute complexity. DP-SGD-RC leverages stochastic trace estimation methods, specifically Hutchinson's estimator[Hutchinson, 1989] and its improved variant, Hutch++[Meyer et al., 2021], to reduce the memory footprint of per-sample gradient norm estimation. We provide a tight privacy analysis showing that DP-SGD-RC achieves noise multipliers competitive with deterministic clipping. Experiments fine-tuning Llama~3.2-1B on long-context benchmarks spanning classification, question answering, and summarization tasks demonstrate that DP-SGD-RC matches baseline utility while significantly reducing memory and compute requirements.

Related papers

• Memory-Efficient Differentially Private Training with Gradient Random Projection [23.309769734156383]
  Differential privacy (DP) protects sensitive data during neural network training.<n>Standard methods like DP-Adam suffer from high memory overhead due to per-sample gradient clipping.<n>We introduce DP-GRAPE (Gradient RAndom ProjEction), a DP training method that significantly reduces memory usage.
  arXiv  Detail & Related papers  (2025-06-18T16:05:09Z)
• Improved Communication-Privacy Trade-offs in $L_2$ Mean Estimation under Streaming Differential Privacy [47.997934291881414]
  Existing mean estimation schemes are usually optimized for $L_infty$ geometry and rely on random rotation or Kashin's representation to adapt to $L$ geometry. We introduce a novel privacy accounting method for the sparsified Gaussian mechanism that incorporates the randomness inherent in sparsification into the DP. Unlike previous approaches, our accounting algorithm directly operates in $L$ geometry, yielding MSEs that fast converge to those of the Gaussian mechanism.
  arXiv  Detail & Related papers  (2024-05-02T03:48:47Z)
• How Private are DP-SGD Implementations? [61.19794019914523]
  We show that there can be a substantial gap between the privacy analysis when using the two types of batch sampling. Our result shows that there can be a substantial gap between the privacy analysis when using the two types of batch sampling.
  arXiv  Detail & Related papers  (2024-03-26T13:02:43Z)
• Differentially Private Zeroth-Order Methods for Scalable Large Language Model Finetuning [0.0]
  differentially private (DP) fine-tuning of pretrained LLMs has been widely used to safeguarding the privacy of task-specific datasets.<n>Despite pushing the scalability of DP-SGD to its limit, DP-SGD-based fine-tuning methods are unfortunately limited by the inherent inefficiency of SGD.
  arXiv  Detail & Related papers  (2024-02-12T17:24:15Z)
• Private Fine-tuning of Large Language Models with Zeroth-order Optimization [51.19403058739522]
  Differentially private gradient descent (DP-SGD) allows models to be trained in a privacy-preserving manner.<n>We introduce DP-ZO, a private fine-tuning framework for large language models by privatizing zeroth order optimization methods.
  arXiv  Detail & Related papers  (2024-01-09T03:53:59Z)
• Differentially Private SGD Without Clipping Bias: An Error-Feedback Approach [62.000948039914135]
  Using Differentially Private Gradient Descent with Gradient Clipping (DPSGD-GC) to ensure Differential Privacy (DP) comes at the cost of model performance degradation. We propose a new error-feedback (EF) DP algorithm as an alternative to DPSGD-GC. We establish an algorithm-specific DP analysis for our proposed algorithm, providing privacy guarantees based on R'enyi DP.
  arXiv  Detail & Related papers  (2023-11-24T17:56:44Z)
• DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Pass [22.578388829171157]
  DP-Forward perturbs embedding in the forward pass of language models. It almost hits the non-private baseline and outperforms DP-SGD by up to 7.7pp at a moderate privacy level.
  arXiv  Detail & Related papers  (2023-09-13T06:37:53Z)
• Normalized/Clipped SGD with Perturbation for Differentially Private Non-Convex Optimization [94.06564567766475]
  DP-SGD and DP-NSGD mitigate the risk of large models memorizing sensitive training data. We show that these two algorithms achieve similar best accuracy while DP-NSGD is comparatively easier to tune than DP-SGD.
  arXiv  Detail & Related papers  (2022-06-27T03:45:02Z)
• Automatic Clipping: Differentially Private Deep Learning Made Easier and Stronger [39.93710312222771]
  Per-example clipping is a key algorithmic step that enables practical differential private (DP) training for deep learning models. We propose an easy-to-use replacement, called automatic clipping, that eliminates the need to tune R for any DPs.
  arXiv  Detail & Related papers  (2022-06-14T19:49:44Z)
• Large Scale Transfer Learning for Differentially Private Image Classification [51.10365553035979]
  Differential Privacy (DP) provides a formal framework for training machine learning models with individual example level privacy. Private training using DP-SGD protects against leakage by injecting noise into individual example gradients. While this result is quite appealing, the computational cost of training large-scale models with DP-SGD is substantially higher than non-private training.
  arXiv  Detail & Related papers  (2022-05-06T01:22:20Z)
• DP-FP: Differentially Private Forward Propagation for Large Models [2.062295244789704]
  We show how to mitigate the performance drop by replacing the Differential Private Gradient Descent with a novel DP Forward-Propagation (DP-FP) Our DP-FP achieves an average accuracy of 91.34% with privacy budgets less than 3, representing a 3.81% performance improvement over the state-of-the-art DP-SGD.
  arXiv  Detail & Related papers  (2021-12-29T07:32:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.