Private Vertical Federated Inference for Time-Series

2026-05-08

AI Summary

This work addresses the challenges of high computational overhead and embedding leakage in privacy-sensitive vertical federated learning (VFL) for sequential data inference. To this end, the authors propose PPHH-VFL, a hybrid architecture that splits the model head into a plaintext public head and a lightweight private head secured via secure multi-party computation (MPC), while incorporating adversarial training to protect the privacy of public embeddings. This approach is the first to simultaneously achieve high efficiency and strong privacy guarantees in vertical federated sequential inference. Experimental results demonstrate that PPHH-VFL accelerates inference by up to one million times compared to end-to-end MPC, and achieves a 44.4× speedup over a VFL+MPC baseline under wide-area network conditions, reducing communication overhead by 91.2% (from 1.7 GB to 19 MB per batch), while improving classification accuracy by 2.50% and lowering regression RMSE by 40.7%.

Abstract

Institutions may benefit from collaborative inference on time-series data. In settings where privacy is necessary, multi-party computation (MPC) is a straightforward approach to providing strong guarantees, yet it remains prohibitively expensive and scales poorly with modern transformer architectures. Vertical Federated Learning (VFL) offers efficiency but suffers from privacy leakage at the embedding level, and securing the entire VFL model head via MPC remains prohibitively slow and communication-heavy for larger models. To enable practical, secure inference at scale, we propose "Public/Private Hybrid Head-VFL" (PPHH-VFL). This hybrid architecture splits the model head into an efficient plaintext public head and a secure, lightweight MPC private head. By applying adversarial training to the public embeddings, we mitigate privacy leakage; concurrently, the small private head securely preserves the flow of sensitive information needed for high downstream utility. Empirical evaluations on models ranging up to 86 million parameters demonstrate that PPHH-VFL accelerates inference by up to six orders of magnitude compared to end-to-end MPC. Compared to a standard VFL+MPC baseline, our approach scales significantly better, achieving a speedup of up to 44.4x in WAN and a 91.2x reduction in communication costs (dropping from 1.7 GB to 19 MB per batch), while simultaneously improving downstream classification accuracy by 2.50% and regression RMSE by 40.7%.

Problem

• Vertical Federated Learning
• Privacy Leakage
• Secure Inference
• Time-Series Data
• Multi-Party Computation

Innovation

• Vertical Federated Learning
• Multi-Party Computation
• Hybrid Architecture
• Adversarial Training
• Privacy-Preserving Inference