EcoFair: Trustworthy and Energy-Aware Routing for Privacy-Preserving Vertically Partitioned Medical Inference

2026-03-27

AI Summary

This work proposes EcoFair, a framework addressing the challenge of simultaneously preserving data locality, ensuring diagnostic reliability, and maintaining edge energy efficiency in privacy-preserving medical inference. Leveraging vertically partitioned learning, EcoFair transmits only modality-specific embeddings to a central server for fusion in dermatological diagnosis. It introduces a lightweight priority routing mechanism that dynamically decides whether to activate a heavyweight image encoder based on predictive uncertainty, the margin between safe and hazardous class probabilities, and a neuro-symbolic risk score. Without altering the global training objective, EcoFair substantially reduces energy consumption at the edge while retaining strong classification performance and significantly enhancing the identification of malignant cases within high-risk subpopulations.

Abstract

Privacy-preserving medical inference must balance data locality, diagnostic reliability, and deployment efficiency. This paper presents EcoFair, a simulated vertically partitioned inference framework for dermatological diagnosis in which raw image and tabular data remain local and only modality-specific embeddings are transmitted for server-side multimodal fusion. EcoFair introduces a lightweight-first routing mechanism that selectively activates a heavier image encoder when local uncertainty or metadata-derived clinical risk indicates that additional computation is warranted. The routing decision combines predictive uncertainty, a safe--danger probability gap, and a tabular neurosymbolic risk score derived from patient age and lesion localisation. Experiments on three dermatology benchmarks show that EcoFair can substantially reduce edge-side inference energy in representative model pairings while remaining competitive in classification performance. The results further indicate that selective routing can improve subgroup-sensitive malignant-case behaviour in representative settings without modifying the global training objective. These findings position EcoFair as a practical framework for privacy-preserving and energy-aware medical inference under edge deployment constraints.

Problem

• privacy-preserving medical inference
• vertically partitioned data
• energy-aware routing
• edge deployment
• multimodal fusion

Innovation

• privacy-preserving inference
• energy-aware routing
• vertically partitioned learning
• multimodal fusion
• neurosymbolic risk scoring