# Removing Disparate Impact of Differentially Private Stochastic Gradient Descent on Model Accuracy

arXiv: https://arxiv.org/abs/2003.03699 · HTML: https://mlsanity.com/paper/2003.03699/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Mar 2020.

## Built on by (most cited first, 13 of 13)

- [Differential Privacy has Bounded Impact on Fairness in Classification](https://mlsanity.com/paper/2210.16242/) (2022)
- [Unlocking Accuracy and Fairness in Differentially Private Image Classification](https://mlsanity.com/paper/2308.10888/) (2023)
- [The Pursuit of Fairness in Artificial Intelligence Models: A Survey](https://mlsanity.com/paper/2403.17333/) (2024)
- [Fair NLP Models with Differentially Private Text Encoders](https://mlsanity.com/paper/2205.06135/) (2022)
- [An Empirical Analysis of Fairness Notions under Differential Privacy](https://mlsanity.com/paper/2302.02910/) (2023)
- [Differential Privacy Under Class Imbalance: Methods and Empirical Insights](https://mlsanity.com/paper/2411.05733/) (2024)
- [Photonic Differential Privacy with Direct Feedback Alignment](https://mlsanity.com/paper/2106.03645/) (2021)
- [On the Importance of Architecture and Feature Selection in Differentially Private Machine Learning](https://mlsanity.com/paper/2205.06720/) (2022)
- [PFGuard: A Generative Framework with Privacy and Fairness Safeguards](https://mlsanity.com/paper/2410.02246/) (2024)
- [Beyond traditional assumptions in fair machine learning](https://mlsanity.com/paper/2101.12476/) (2021)
- [Combining Machine Learning Defenses without Conflicts](https://mlsanity.com/paper/2411.09776/) (2024)
- [SoftAdaClip: A Smooth Clipping Strategy for Fair and Private Model Training](https://mlsanity.com/paper/2510.01447/) (2025)
- [Towards Reliable and Generalizable Differentially Private Machine Learning (Extended Version)](https://mlsanity.com/paper/2508.15141/) (2025)

## Abstract (arXiv)

When we enforce differential privacy in machine learning, the utility-privacy trade-off is different w.r.t. each group. Gradient clipping and random noise addition disproportionately affect underrepresented and complex classes and subgroups, which results in inequality in utility loss. In this work, we analyze the inequality in utility loss by differential privacy and propose a modified differentially private stochastic gradient descent (DPSGD), called DPSGD-F, to remove the potential disparate impact of differential privacy on the protected group. DPSGD-F adjusts the contribution of samples in a group depending on the group clipping bias such that differential privacy has no disparate impact on group utility. Our experimental evaluation shows how group sample size and group clipping bias affect the impact of differential privacy in DPSGD, and how adaptive clipping for each group helps to mitigate the disparate impact caused by differential privacy in DPSGD-F.