Toward Individual Fairness Without Centralized Data: Selective Counterfactual Consistency for Vertical Federated Learning

2026-05-07

AI Summary

This work addresses the challenge of achieving individual-level fairness in vertical federated learning (VFL), where features are distributed across multiple parties and sensitive attributes remain private. The authors propose SCC-VFL, a novel framework that, for the first time, enables counterfactual fairness without centralized data access. SCC-VFL leverages differential privacy to identify feature roles, selectively edits only the mediators to generate masked counterfactuals, and introduces a server-side consistency loss to ensure prediction stability under interventions on protected attributes—all while preserving privacy. Empirical evaluations on credit, healthcare, and criminal justice datasets demonstrate that the method reduces decision flip rates by up to 98%, maintains or improves predictive accuracy, and effectively resists attribute inference attacks, thereby enhancing model robustness.

Abstract

When algorithmic decisions depend on data distributed across institutions, how can we ensure that an individual's outcome does not change arbitrarily based on a protected attribute? We study this question in vertical federated learning (VFL), where features are split across parties, sensitive attributes may be private, and proxies for protected characteristics can be scattered across institutional boundaries under strict privacy constraints. Our focus is on individual-level counterfactual stability, i.e., per-instance prediction consistency under protected-attribute interventions as formalized in the causal fairness literature, rather than group parity guarantees such as demographic parity or equalized odds. We propose SCC-VFL, a server-centric framework for enforcing selective counterfactual consistency (SCC) at the individual level in VFL. SCC-VFL operationalizes a given policy specification by combining three components: (i) differentially private, graph-free discovery of feature roles into non-descendants, policy-permitted mediators, and impermissible proxies using only a formally private sketch of the sensitive attribute, with a formal per-release privacy that does not extend to the full training pipeline; (ii) masked counterfactual generation that edits only mediators while fixing non-descendants and suppressing proxy leakage; and (iii) server-side enforcement via an SCC consistency loss that penalizes impermissible prediction changes under protected-attribute interventions. Across three real-world datasets spanning credit, healthcare, and criminal justice, SCC-VFL maintains or improves predictive accuracy while sharply reducing decision flip rates by up to 98% relative to strong baselines. It also lowers attribute-inference attack success and improves robustness, demonstrating favorable utility-fairness-privacy trade-offs in realistic VFL deployments.

Problem

• individual fairness
• vertical federated learning
• counterfactual consistency
• protected attribute
• privacy

Innovation

• Selective Counterfactual Consistency
• Vertical Federated Learning
• Individual Fairness
• Differentially Private Feature Role Discovery
• Causal Fairness