We are delighted that Prof. Allen AU, our Professor; Dr Xingye LU, our Research Assistant Professor; Mengling LIU, our PhD student; and Chengru ZHANG, PhD student at HKU, have won the Plonk-DIZK GPU Acceleration Prize in the ZPrize Competition, receiving an award of $550,000 USD (approximately HKD 4,300,000).
Zero-knowledge proofs (ZKPs), which allow one to prove the validity of an assertion without revealing any other information, are considered a game-changing technology for Web3 as they can be used to protect data privacy and improve the efficiency, and scalability of transparent systems such as blockchain. It has the potential to be applied across a variety of industries, including financial services, healthcare, and supply chain management. However, a major challenge for the real-world adoption of ZKPs lies in their computational cost, particularly for applications at scale. To overcome this hurdle, the industry-wide ZPrize Competition was launched in 2022, aiming to enhance the performance of zero-knowledge proofs (ZKPs) for different applications or hardware platforms, which can significantly revolutionise user experience regarding privacy, security, and integrity, especially for blockchain applications. Teams competed for monetary prizes spanning a range of categories (7 open divisions and 5 team divisions). Over $7M in prize money has been committed.
The Plonk-DIZK GPU Acceleration category focused on accelerating a specific type of ZKPs called PLONK, which can be parallelised by distributing the proof generation procedure among multiple servers. Participants were required to apply the Distributed Zero-Knowledge Proof (DIZK) approach to accelerate the Plonk zero-knowledge proof protocol and improve the efficiency of Plonk-proof generation for gigantic circuits. However, due to the insufficient memory capacity of servers and some other real-world factors, it is considered to be extremely difficult to perform the acceleration.
As a result, only this team led by Prof. Au and Dr Xingye Lu successfully completed the tasks by demonstrating a 40% computation speedup compared to the baseline on a single machine. Compared with the original DIZK, which requires more than 2 hours with twenty powerful servers (each equipped with 244GB of RAM) to prove a circuit with 228 gates, the team only needs 48 minutes to generate such a proof. The team's approach involved moving the most time-consuming operations in the proof generation process to GPUs and using a categorisation of servers into dispatchers and workers to distribute the prover's computation across a cluster of computers with minimised communication costs.
Their solution is believed to be crucial for the broader adoption of ZKPs in various applications as it optimises scalability, efficiency, and resource consumption. The team will further generalise their distribution approach on PLONK and apply it to other ZKPs.
Overall, the ZPrize Competition plays a significant role in advancing the field of zero-knowledge proof. We congratulate the team on their impressive achievement and look forward to seeing the impact of their winning submission on the broader adoption of zero-knowledge proof in the future.
Zero-knowledge proofs (ZKPs), which allow one to prove the validity of an assertion without revealing any other information, are considered a game-changing technology for Web3 as they can be used to protect data privacy and improve the efficiency, and scalability of transparent systems such as blockchain. It has the potential to be applied across a variety of industries, including financial services, healthcare, and supply chain management. However, a major challenge for the real-world adoption of ZKPs lies in their computational cost, particularly for applications at scale. To overcome this hurdle, the industry-wide ZPrize Competition was launched in 2022, aiming to enhance the performance of zero-knowledge proofs (ZKPs) for different applications or hardware platforms, which can significantly revolutionise user experience regarding privacy, security, and integrity, especially for blockchain applications. Teams competed for monetary prizes spanning a range of categories (7 open divisions and 5 team divisions). Over $7M in prize money has been committed.
The Plonk-DIZK GPU Acceleration category focused on accelerating a specific type of ZKPs called PLONK, which can be parallelised by distributing the proof generation procedure among multiple servers. Participants were required to apply the Distributed Zero-Knowledge Proof (DIZK) approach to accelerate the Plonk zero-knowledge proof protocol and improve the efficiency of Plonk-proof generation for gigantic circuits. However, due to the insufficient memory capacity of servers and some other real-world factors, it is considered to be extremely difficult to perform the acceleration.
As a result, only this team led by Prof. Au and Dr Xingye Lu successfully completed the tasks by demonstrating a 40% computation speedup compared to the baseline on a single machine. Compared with the original DIZK, which requires more than 2 hours with twenty powerful servers (each equipped with 244GB of RAM) to prove a circuit with 228 gates, the team only needs 48 minutes to generate such a proof. The team's approach involved moving the most time-consuming operations in the proof generation process to GPUs and using a categorisation of servers into dispatchers and workers to distribute the prover's computation across a cluster of computers with minimised communication costs.
Their solution is believed to be crucial for the broader adoption of ZKPs in various applications as it optimises scalability, efficiency, and resource consumption. The team will further generalise their distribution approach on PLONK and apply it to other ZKPs.
Overall, the ZPrize Competition plays a significant role in advancing the field of zero-knowledge proof. We congratulate the team on their impressive achievement and look forward to seeing the impact of their winning submission on the broader adoption of zero-knowledge proof in the future.
