| Title |
Research Analysis and Future Directions for Verifiable Federated Learning |
| Authors |
김시온(Sion Kim) ; 유동현(Donghyun Yu) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.7.83 |
| Keywords |
Federated learning; Verifiable federated learning; Aggregation integrity; Proof of learning |
| Abstract |
Federated Learning (FL) is a privacy-preserving paradigm that enables collaborative model training in decentralized environments without data movement. However, the risk of information leakage through updates and integrity threats from malicious clients have emerged as core challenges, creating a structural trade-off between privacy protection and model reliability. To overcome these limitations, Verifiable Federated Learning (VFL) has been proposed. VFL leverages techniques such as Zero-Knowledge Proofs (ZKPs) and Trusted Execution Environments (TEE) to verify the correctness of training and aggregation without exposing private data. This paper systematically analyzes privacy and integrity threats in FL and categorizes existing VFL research into three classes based on the verification subject and scope: server-side aggregation verification, client-side training verification, and integrated approaches. We further analyze the structural limitations of VFL, focusing on computational costs and scalability issues arising from expanded verification scopes, and suggest future research directions. |