# Empirical Calibration and Metric Differential Privacy in Language Models

arXiv: https://arxiv.org/abs/2503.13872 · HTML: https://mlsanity.com/paper/2503.13872/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 53 of its 91 references that are in the map.
- First posted on arXiv 18 Mar 2025.

## Builds on (most cited first, 30 of 53)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Deep Learning with Differential Privacy](https://mlsanity.com/paper/1607.00133/) (2016)
- [Deep Leakage from Gradients](https://mlsanity.com/paper/1906.08935/) (2019)
- [Neural Network Acceptability Judgments](https://mlsanity.com/paper/1805.12471/) (2018)
- [Inverting Gradients -- How easy is it to break privacy in federated learning?](https://mlsanity.com/paper/2003.14053/) (2020)
- [Renyi Differential Privacy](https://mlsanity.com/paper/1702.07476/) (2017)
- [Large Language Models Can Be Strong Differentially Private Learners](https://mlsanity.com/paper/2110.05679/) (2021)
- [Label-Only Membership Inference Attacks](https://mlsanity.com/paper/2007.14321/) (2020)
- [iDLG: Improved Deep Leakage from Gradients](https://mlsanity.com/paper/2001.02610/) (2020)
- [Concentrated Differential Privacy: Simplifications, Extensions, and Lower Bounds](https://mlsanity.com/paper/1605.02065/) (2016)
- [Membership Inference Attacks on Machine Learning: A Survey](https://mlsanity.com/paper/2103.07853/) (2021)
- [Membership Inference Attacks against Language Models via Neighbourhood Comparison](https://mlsanity.com/paper/2305.18462/) (2023)
- [Enhanced Membership Inference Attacks against Machine Learning Models](https://mlsanity.com/paper/2111.09679/) (2021)
- [Differentially Private Learning Needs Better Features (or Much More Data)](https://mlsanity.com/paper/2011.11660/) (2020)
- [Unlocking High-Accuracy Differentially Private Image Classification through Scale](https://mlsanity.com/paper/2204.13650/) (2022)
- [Auditing Differentially Private Machine Learning: How Private is Private SGD?](https://mlsanity.com/paper/2006.07709/) (2020)
- [Trustworthy AI: From Principles to Practices](https://mlsanity.com/paper/2110.01167/) (2021)
- [Do Membership Inference Attacks Work on Large Language Models?](https://mlsanity.com/paper/2402.07841/) (2024)
- [Quantifying Privacy Risks of Masked Language Models Using Membership Inference Attacks](https://mlsanity.com/paper/2203.03929/) (2022)
- [How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy](https://mlsanity.com/paper/2303.00654/) (2023)
- [On the Importance of Difficulty Calibration in Membership Inference Attacks](https://mlsanity.com/paper/2111.08440/) (2021)
- [Concentrated Differential Privacy](https://mlsanity.com/paper/1603.01887/) (2016)
- [Privacy- and Utility-Preserving Textual Analysis via Calibrated Multivariate Perturbations](https://mlsanity.com/paper/1910.08902/) (2019)
- [Large-Scale Differentially Private BERT](https://mlsanity.com/paper/2108.01624/) (2021)
- [Robbing the Fed: Directly Obtaining Private Data in Federated Learning with Modified Models](https://mlsanity.com/paper/2110.13057/) (2021)
- [Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning](https://mlsanity.com/paper/1812.00910/) (2018)
- [Differential Privacy for Text Analytics via Natural Text Sanitization](https://mlsanity.com/paper/2106.01221/) (2021)
- [ML Privacy Meter: Aiding Regulatory Compliance by Quantifying the Privacy Risks of Machine Learning](https://mlsanity.com/paper/2007.09339/) (2020)
- [Recovering Private Text in Federated Learning of Language Models](https://mlsanity.com/paper/2205.08514/) (2022)
- [TAG: Gradient Attack on Transformer-based Language Models](https://mlsanity.com/paper/2103.06819/) (2021)

## Built on by (most cited first, 1 of 1)

- [Beyond Theoretical Bounds: Empirical Privacy Loss Calibration for Text Rewriting Under Local Differential Privacy](https://mlsanity.com/paper/2603.22968/) (2026)

## Abstract (arXiv)

NLP models trained with differential privacy (DP) usually adopt the DP-SGD framework, and privacy guarantees are often reported in terms of the privacy budget ε. However, ε does not have any intrinsic meaning, and it is generally not possible to compare across variants of the framework. Work in image processing has therefore explored how to empirically calibrate noise across frameworks using Membership Inference Attacks (MIAs). However, this kind of calibration has not been established for NLP. In this paper, we show that MIAs offer little help in calibrating privacy, whereas reconstruction attacks are more useful. As a use case, we define a novel kind of directional privacy based on the von Mises-Fisher (VMF) distribution, a metric DP mechanism that perturbs angular distance rather than adding (isotropic) Gaussian noise, and apply this to NLP architectures. We show that, even though formal guarantees are incomparable, empirical privacy calibration reveals that each mechanism has different areas of strength with respect to utility-privacy trade-offs.