🤖 AI Summary
To address the dual challenges of privacy leakage (e.g., gradient inversion attacks) and high communication overhead in medical federated learning, this paper proposes a privacy-enhancing framework integrating Vision Transformers (ViT) with CKKS homomorphic encryption. The core method replaces conventional encrypted gradient aggregation with homomorphically encrypted CLS tokens from ViT—enabling secure, feature-level aggregation and direct inference directly in the ciphertext domain. This design eliminates gradient reconstruction vulnerabilities while drastically reducing communication costs: per-round overhead drops by 30× to merely 326 KB. Evaluated on a lung cancer histopathological classification task, the framework achieves a global accuracy of 96.12%, with ciphertext-domain inference maintaining 90.02% accuracy. Thus, it simultaneously delivers strong privacy guarantees, substantial communication efficiency gains, and competitive model performance.
📝 Abstract
Collaborative machine learning across healthcare institutions promises improved diagnostic accuracy by leveraging diverse datasets, yet privacy regulations such as HIPAA prohibit direct patient data sharing. While federated learning (FL) enables decentralized training without raw data exchange, recent studies show that model gradients in conventional FL remain vulnerable to reconstruction attacks, potentially exposing sensitive medical information. This paper presents a privacy-preserving federated learning framework combining Vision Transformers (ViT) with homomorphic encryption (HE) for secure multi-institutional histopathology classification. The approach leverages the ViT CLS token as a compact 768-dimensional feature representation for secure aggregation, encrypting these tokens using CKKS homomorphic encryption before transmission to the server. We demonstrate that encrypting CLS tokens achieves a 30-fold communication reduction compared to gradient encryption while maintaining strong privacy guarantees. Through evaluation on a three-client federated setup for lung cancer histopathology classification, we show that gradients are highly susceptible to model inversion attacks (PSNR: 52.26 dB, SSIM: 0.999, NMI: 0.741), enabling near-perfect image reconstruction. In contrast, the proposed CLS-protected HE approach prevents such attacks while enabling encrypted inference directly on ciphertexts, requiring only 326 KB of encrypted data transmission per aggregation round. The framework achieves 96.12 percent global classification accuracy in the unencrypted domain and 90.02 percent in the encrypted domain.