[2502.18535] A Survey of Zero-Knowledge Proof Based Verifiable Machine Learning

Computer Science > Cryptography and Security arXiv:2502.18535 (cs) [Submitted on 25 Feb 2025 (v1), last revised 29 Mar 2026 (this version, v2)] Title:A Survey of Zero-Knowledge Proof Based Verifiable Machine Learning Authors:Zhizhi Peng, Chonghe Zhao, Taotao Wang, Guofu Liao, Zibin Lin, Yifeng Liu, Bin Cao, Long Shi, Qing Yang, Shengli Zhang View a PDF of the paper titled A Survey of Zero-Knowledge Proof Based Verifiable Machine Learning, by Zhizhi Peng and 9 other authors View PDF HTML (experimental) Abstract:Machine learning is increasingly deployed through outsourced and cloud-based pipelines, which improve accessibility but also raise concerns about computational integrity, data privacy, and model confidentiality. Zero-knowledge proofs (ZKPs) provide a compelling foundation for verifiable machine learning because they allow one party to certify that a training, testing, or inference result was produced by the claimed computation without revealing sensitive data or proprietary model parameters. Despite rapid progress in zero-knowledge machine learning (ZKML), the literature remains fragmented across different cryptographic settings, ML tasks, and system objectives. This survey presents a comprehensive review of ZKML research published from June 2017 to August 2025. We first introduce the basic ZKP formulations underlying ZKML and organize existing studies into three core tasks: verifiable training, verifiable testing, and verifiable inference. We then synthesize represe...