Blockchain holds the potential to support the privacy of its participants by integrating zero-knowledge proofs into its design. zkSNARK schemes can effectively prove the validity of blockchain transactions to their owners without disclosing user data. However, the efficiency and scalability of the underlying cryptographic protocols that enable these schemes remain a challenge to realize in practice. This paper presents Air-FRI, a novel GPU-enabled software implementation of the Fast Reed-Solomon Interactive Oracle Proof of Proximity (FRI) protocol, which is a core component in post-quantum zkSNARKs that reduces computational complexity in systems with substantial mathematical instances. Existing schemes that implement the FRI protocol entail significant computational times due to large proof sizes. Our optimized solution includes a parallelized computation of Reed-Solomon codewords, the pre-computation of time-intensive finite field operations, non-interactiveness, and the application of a unified Merkle tree commitment to authenticate the entire proof. Together, the implemented optimizations yield a solution that significantly reduces prover and verifier times while minimizing proof size, addressing both scalability and performance challenges in zkSNARK-based algorithms. Performance evaluations conducted by us across two security levels confirm the implementation’s high throughput, establishing it as a promising solution for practicable privacy-preservation. Our results show a 93.3% improvement on an average in the speed of the protocol on the GPU as compared to a non-GPU solution for the same parameters, which includes the execution time of all of its sub-phases: the commit phase, the query phase, and the round consistency checks to ensure correctness of the proof. This work establishes a foundation for advancing privacy-preserving post-quantum cryptography, supporting the development of secure, transparent, and decentralized digital infrastructures for the future.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Air-FRI: Acceleration of the FRI Protocol on the GPU for ZkSNARK Applications

  • Tanmayi Jandhyala,
  • Guang Gong

摘要

Blockchain holds the potential to support the privacy of its participants by integrating zero-knowledge proofs into its design. zkSNARK schemes can effectively prove the validity of blockchain transactions to their owners without disclosing user data. However, the efficiency and scalability of the underlying cryptographic protocols that enable these schemes remain a challenge to realize in practice. This paper presents Air-FRI, a novel GPU-enabled software implementation of the Fast Reed-Solomon Interactive Oracle Proof of Proximity (FRI) protocol, which is a core component in post-quantum zkSNARKs that reduces computational complexity in systems with substantial mathematical instances. Existing schemes that implement the FRI protocol entail significant computational times due to large proof sizes. Our optimized solution includes a parallelized computation of Reed-Solomon codewords, the pre-computation of time-intensive finite field operations, non-interactiveness, and the application of a unified Merkle tree commitment to authenticate the entire proof. Together, the implemented optimizations yield a solution that significantly reduces prover and verifier times while minimizing proof size, addressing both scalability and performance challenges in zkSNARK-based algorithms. Performance evaluations conducted by us across two security levels confirm the implementation’s high throughput, establishing it as a promising solution for practicable privacy-preservation. Our results show a 93.3% improvement on an average in the speed of the protocol on the GPU as compared to a non-GPU solution for the same parameters, which includes the execution time of all of its sub-phases: the commit phase, the query phase, and the round consistency checks to ensure correctness of the proof. This work establishes a foundation for advancing privacy-preserving post-quantum cryptography, supporting the development of secure, transparent, and decentralized digital infrastructures for the future.