Machine-checking Multi-Round Proofs of Shuffle: Terelius-Wikstrom and Bayer-Groth by Thomas Haines, Rajeev Goré and Mukesh Tiwari: https://eprint.iacr.org/2025/461

This paper discusses the machine verification of the complex Bayer-Groth proof of shuffle using the Coq proof assistant, resulting in a verified verifier compatible with real-world e-voting systems like Swiss Post’s. This proof is critical in ensuring ballot privacy and integrity in electronic elections. The work improves on earlier efforts by formally checking completeness, soundness, and ZK properties, and confirming that the Swiss implementation conforms to the mathematical design. It bridges a longstanding gap between theory and secure software practice.

zkAML: Zero-knowledge Anti Money Laundering in Smart Contracts with whitelist approach by Donghwan Oh, Semin Han, Jihye Kim, Hyunok Oh, Jiyeal Chung, Jieun Lee, Hee-jun Yoo and Tae wan Kim: https://eprint.iacr.org/2025/465

This paper presents zkAML, a cryptographic framework aimed at improving anti-money laundering (AML) compliance in smart contracts. By using ZKPs, the system allows users to prove regulatory compliance without revealing personal data. This reduces redundant identity checks, preserves privacy, and enhances efficiency. Tested on blockchain networks, zkAML reached up to 324 transactions per second, showing potential for real-world apps like digital currencies and cross-border transfers.

HammR: A ZKP Protocol for Fixed Hamming-Weight Restricted-Entry Vectors by Felice Manganiello and Freeman Slaughter: https://eprint.iacr.org/2025/475

This paper introduces HammR, a ZKP protocol that allows a prover to demonstrate knowledge of a vector with fixed Hamming weight and restricted entries, without revealing the vector itself. The protocol uses inner product arguments and Pedersen commitments to ensure ZK, soundness, and completeness. HammR is flexible, supporting batched proofs and non-interactive forms via Fiat-Shamir. It is applicable to various syndrome decoding problems and broader settings like lookup instances, proximity proofs, and secure electronic voting.

PREAMBLE: Private and Efficient Aggregation of Block Sparse Vectors and Applications by Hilal Asi, Vitaly Feldman, Hannah Keller, Guy N. Rothblum and Kunal Talwar: https://eprint.iacr.org/2025/490

This paper proposes PREAMBLE, a communication- and computation-efficient method for aggregating block-sparse vectors in secure two-server systems like Prio. Designed for high-dimensional data, PREAMBLE significantly reduces communication costs in private federated learning by efficiently secret-sharing clustered non-zero vector segments. It supports privacy-preserving aggregation with minimal impact on noise variance. For instance, aggregating eight-million-dimensional vectors requires only ~1MB of communication per client, compared to 64MB in prior systems, while maintaining near-optimal differential privacy guarantees.

Endorser Peer Anonymization in Hyperledger Fabric for Consortium of Organizations by Dharani J, Sundarakantham K, Kunwar Singh and Mercy Shalinie S: https://eprint.iacr.org/2025/492

This paper proposes a privacy-preserving endorsement system for Hyperledger Fabric to protect the identity of endorsing organizations in a blockchain consortium. The approach uses a scoped-linkable threshold ring signature to anonymize endorsers and secures endorsement policies through Pedersen commitments and ZKPs. It maintains both efficiency and security, addressing known vulnerabilities in Fabric related to endorser and policy exposure.

Scalable Zero-knowledge Proofs for Non-linear Functions in Machine Learning by Meng Hao, Hanxiao Chen, Hongwei Li, Chenkai Weng, Yuan Zhang, Haomiao Yang and Tianwei Zhang: https://eprint.iacr.org/2025/507

The paper introduces a scalable ZKP framework aimed at efficiently verifying non-linear functions in machine learning models. By rethinking ZKP construction through table lookup techniques, the authors reduce the significant computational overhead caused by evaluating non-linear layers. The framework includes novel building blocks like digital decomposition and comparison, enabling fast and sound ZKPs for functions such as ReLU and sigmoid. Experimental results show a 50–179× speed improvement over prior methods, while keeping communication costs low.

Server-Aided Anonymous Credentials by Rutchathon Chairattana-Apirom, Franklin Harding, Anna Lysyanskaya and Stefano Tessaro: https://eprint.iacr.org/2025/513

This paper describes the concept of Server-Aided Anonymous Credentials, a model that allows users to prove possession of credentials with the help of a server, without revealing their identity or attributes. The approach relies on ZKPs to maintain unlinkability and privacy, even when server assistance is used. The authors propose concrete constructions using pairing-free cryptography, making the system more compatible with standardized and widely supported elliptic curves.

On Extractability of the KZG Family of Polynomial Commitment Schemes by Juraj Belohorec, Pavel Dvořák, Charlotte Hoffmann, Pavel Hubáček, Kristýna Mašková and Martin Pastyřík: https://eprint.iacr.org/2025/514

This paper presents a unifying framework for analyzing knowledge-soundness in KZG polynomial commitment schemes, expanding it to both univariate and multivariate forms. It defines a new Proof-of-Knowledge of a Polynomial to formalize extractability, crucial for zk-SNARKs. Using a novel decomposition lemma and generalized cryptographic assumptions, the authors establish the first standard-model extractability proofs for multivariate KZG schemes. Their results extend to several KZG variants and offer clearer foundations for secure, efficient ZK protocols without relying on idealized models.

Compressed Sigma Protocols: New Model and Aggregation Techniques by Yuxi Xue, Tianyu Zheng, Shang Gao, Bin Xiao and Man Ho Au: https://eprint.iacr.org/2025/515

This study examines a more flexible and compact approach to Sigma protocols, commonly used in cryptographic proofs. The new model introduces a structure called “doubly folded commitments” that can better handle complex, non-linear constraints. It also supports an aggregation technique that reduces proof sizes. This improvement allows for more efficient implementation of privacy-preserving technologies like ring signatures and k-out-of-n proofs without relying on costly cryptographic tools like pairing-friendly curves.

Masking-Friendly Post-Quantum Signatures in the Threshold-Computation-in-the-Head Framework by Thibauld Feneuil, Matthieu Rivain and Auguste Warmé-Janville: https://eprint.iacr.org/2025/520

This paper presents new methods for securing post-quantum signature schemes against side-channel attacks using the Threshold-Computation-in-the-Head framework. The authors analyze vulnerabilities in existing schemes and propose masking techniques to enhance security, introducing three specific tweaks that significantly reduce computational overhead. Their approach enables improved efficiency, especially for higher masking orders, and offers practical trade-offs between performance and signature size. Benchmarks on RISC-V platforms show notable speedups over traditional masked implementations, making these techniques promising for embedded post-quantum cryptography.

VeRange: Verification-efficient Zero-knowledge Range Arguments with Transparent Setup for Blockchain Applications and More by Yue Zhou and Sid Chi-Kin Chau: https://eprint.iacr.org/2025/528

The paper introduces VeRange, a new suite of ZK range arguments designed to significantly reduce verification costs on blockchains. Unlike existing methods, VeRange operates with a transparent setup and achieves lower gas fees and faster verification by minimizing group exponentiations. It supports aggregating multiple range proofs into one, enhancing scalability. Empirical deployment on Ethereum confirms VeRange as the most verification-efficient among discrete-logarithm-based range arguments, making it especially practical for decentralized apps like confidential transactions, solvency proofs, and anonymous credentials.

JesseQ: Efficient Zero-Knowledge Proofs for Circuits over Any Field by Mengling Liu, Yang Heng, Xingye Lu and Man Ho Au: https://eprint.iacr.org/2025/533

This paper reveals JesseQ, a suite of two efficient ZKP protocols - JQv1 and JQv2 - based on VOLE. These protocols support circuits over any field and offer significant improvements in prover speed and communication cost. Experiments show up to 7× faster performance for Boolean circuits compared to prior systems like QuickSilver, with practical efficiency for tasks like inner products, matrix multiplication, and lattice problem proofs. JesseQ also integrates well with sublinear frameworks for batched disjunctive statements.

Plonkify: R1CS-to-Plonk transpiler by Pengfei Zhu: https://eprint.iacr.org/2025/534

This paper presents Plonkify - a tool that translates circuits from the R1CS format to the Plonk constraint system while aiming to minimize constraint overhead. It supports both vanilla and custom Plonk gates, offering notable improvements over existing converters. On a benchmark circuit, Plonkify reduced constraints from over 2 million to under 860,000, making it a practical option for developers looking to transition to more efficient Plonk-based zkSNARK systems.

zkPyTorch: A Hierarchical Optimized Compiler for Zero-Knowledge Machine Learning by Tiancheng Xie, Tao Lu, Zhiyong Fang, Siqi Wang, Zhenfei Zhang, Yongzheng Jia, Dawn Song and Jiaheng Zhang: https://eprint.iacr.org/2025/535

This work presents ZKPyTorch, a compiler that connects PyTorch-based machine learning models with ZKP engines. It automates the conversion of ML models into proof-compatible circuits through preprocessing, ZKP-friendly quantization, and hierarchical optimization. This allows AI developers to generate verifiable, privacy-preserving computations without cryptographic expertise. The approach was tested on VGG-16 and Llama-3, showing strong performance and paving the way for practical adoption of zero-knowledge machine learning.

‍