Skip to content

Pushing the (Decision) Boundaries: Dynamically Calibrating Differentially Private Noise to Explainability in Federated Learning

Sep 2026 · 0 citations · 49 references
Computer Science

TL;DR

XCal-FL is proposed, a closed-loop, explainability-driven local training algorithm for image classification in cross-silo FL that dynamically calibrates DP noise from three complementary signals, suggesting explainability is a distinct dimension of the privacy trade-off that cannot be inferred from utility alone, with implications for training and privacy-budget allocation in decision-critical applications.

Abstract

Federated Learning (FL) with Differential Privacy (DP) is increasingly adopted to preserve data confidentiality in distributed machine learning. However, DP noise distorts learned representations and degrades explanation fidelity, limiting differentially private FL where trustworthy explanations are required, such as assistive clinical diagnosis. Prior work adapted DP noise with static feature-importance signals, restricting explainability to post hoc analysis and precluding noise calibration to explanation quality during training. We propose XCal-FL, a closed-loop, explainability-driven local training algorithm for image classification in cross-silo FL that dynamically calibrates DP noise from three complementary signals: (1) prediction logit variations, measuring causal influence on model confidence, (2) counterfactual margins, capturing decision-boundary sensitivity, and (3) saliency concentration, quantifying spatial coherence of model attention, while enforcing formal DP guarantees via adaptive privacy accounting. Experiments on three medical imaging datasets across varying FL configurations show that XCal-FL yields more accurate and interpretable global models, improving predictive performance by over 10\% and explanation fidelity by up to 5$\times$ over static-noise FL, and outperforming state-of-the-art adaptive DP methods in fidelity. XCal-FL also achieves higher privacy-budget efficiency, turning each unit of cumulative privacy loss into larger gains in both accuracy and explanation fidelity. Our analysis further reveals that, unlike predictive performance, which scales roughly linearly with privacy loss, explanation fidelity exhibits non-linear dynamics. These findings suggest explainability is a distinct dimension of the privacy trade-off that cannot be inferred from utility alone, with implications for training and privacy-budget allocation in decision-critical applications.

View source

Similar papers

#federated learning Open access Sep 2026

Understanding Differential Privacy in Decentralized Federated Learning: A Controlled Privacy–Utility Comparison

Results show that clipping, perturbation, and the learning procedure jointly influence the observed privacy–utility behavior of centralized FL and sequential Decentralized Federated Learning.

AlsharifHasan Mohamad Aburbeian, M. Fernández-Veiga, A. Fernández-Vilas et al. · 0 citations

Efficient Active Auditing of Multi-Group Fairness with Bias Probes

Over the past decade, Machine Learning (ML) has been trained under dual objectives: minimizing prediction error via Empirical Risk Minimization (ERM) while controlling unfairness bias. In practice, however, fairness-aware training often yields limited improvements over standard ERM, making reliable post hoc auditing es...

Ayoub Ajarra, Debabrota Basu · 0 citations
Open access Sep 2026

Research on a Quantitative Evaluation Model for Privacy Computing Technology Selection

This work comparatively examines FL and MPC across credit scoring and medical image segmentation tasks, finding that MPC delivers robust security at a punitive cost, rendering it economically fragile for deep learning pipelines, whereas FL proves resilient under data heterogeneity yet remains exposed by its semi-honest...

Cheng-Ming Li · 0 citations

Principled Bias Detection and Mitigation across the ML Data Lifecycle

This dissertation treats data bias as a first-class data management problem and develops principled frameworks for detecting and correcting it across the ML data lifecycle, introducing Uniform Bias (UB), an interpretable intersectional measure with formal guarantees, and an ILP-based mitigation framework that models co...

Unknown authors · 0 citations
2026

Decision Boundary Drift: A Security, Privacy, and Trust Risk of Continual Learning for Agentic AI in Edge Networks

Continual learning (CL) is a key paradigm that enables intelligent agents to operate autonomously in edge networks over the long term. However, continuous model updates can lead to catastrophic forgetting and representation instability in edge deployment scenarios, which may further induce Decision Boundary Drift (DBD)...

Kai-Xiang Yang, Yue-Bin Xu, Zhi-Hao Li et al. · 0 citations
Open access Aug 2026

Regulatory-Driven Federated Learning: A Multi-Objective Approach to Compliance and Ethical AI in Financial Systems

A regulatory-driven FL framework that treats compliance and fairness as first-class optimization objectives rather than as afterthoughts is proposed, and the results surface a known tension: enforcing demographic parity increases the equalized-odds gap, quantifying the price of fairness under differential privacy in fe...

J. Nalavade · 0 citations

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.