Selected Publications:
SoK: Blockchain Agent-to-Agent Payments at Blockchain: Research and Applications (BCRA, JCR Q1)
Yuanzhe Zhang, Yuexin Xiang, Yuchen Lei, et al.
This work systematizes blockchain-based payment infrastructures for Agent-to-Agent (A2A) ecosystems. We study A2A payments through a four-stage lifecycle of discovery, authorization, execution, and accounting, and analyze how existing mechanisms support or fail to support trustworthy autonomous payment interactions. The study identifies key trust and security challenges, including weak intent binding, misuse under valid authorization, payment-service decoupling, and limited accountability. It further argues that A2A payment systems require end-to-end compositional correctness, where the user’s intent, authorized action, executed payment, and delivered service outcome remain consistently aligned across the full payment lifecycle.
Mosaic: Client-driven Account Allocation Framework in Sharded Blockchains at the 45th IEEE International Conference on Distributed Computing Systems (ICDCS 2025, CORE ranking-A)
Yuanzhe Zhang, Shirui Pan, Jiangshan Yu.
Recent account allocation studies in sharded blockchains are typically miner-driven, requiring miners to perform global optimizations for all accounts to enhance system- wide performance. This forces each miner to maintain a complete copy of the entire ledger, resulting in significant storage, communication, and computation overhead. In this work, we explore an alternative research direction by proposing Mosaic, the first client-driven framework for distributed, lightweight local optimization. Rather than relying on miners to allocate all accounts, Mosaic enables clients to independently execute a local algorithm to determine their residing shards. Mosaic naturally addresses key limitations of miner-driven approaches, including the lack of miner incentives and the significant overhead. While clients are flexible to adopt any algorithm for shard allocation, we design and implement a reference algorithm, Pilot, to guide them. Clients execute Pilot to maximize their own benefits, such as reduced transaction fees and confirmation latency. We implement and evaluate Pilot against state-of-the-art miner-driven global optimization solutions. The results demonstrate that Mosaic significantly enhances computational efficiency, achieving a four-order-of- magnitude reduction in computation time, with the reduced input data size from 1.44 GB to an average of 228.66 bytes per account. Despite these efficiency gains, Pilot introduces only about a 5% increase in the cross-shard ratio and maintains approximately 98% of the system throughput, demonstrating a minimal trade- off in overall effectiveness.
TxAllo: Dynamic Transaction Allocation in Sharded Blockchain Systems at the 39th IEEE International Conference on Data Engineering (ICDE 2023, CORE ranking-A*)
Yuanzhe Zhang, Shirui Pan, Jiangshan Yu.
This work investigated the transaction and account allocation problem in sharded blockchains to reduce the expensive cross-shard transactions. We formulated and converted this problem and key concepts from the blockchain domain to the graph domain and proposed a community detection algorithm, TxAllo, to optimize the throughput on this graph. TxAllo simultaneously considered the number of cross-shard transactions and the workload balance among shards. The experiments illustrate that the ratio of cross-shard transactions can be significantly reduced from about 98% to 12%. In the meantime, the workload balance is also well maintained in different shards. For the adaptive updating method A-TxAllo, the execution time is approximately 800 times faster.
Correlation-aware Next Basket Recommendation using Graph Attention Networks at the 27th International Conference on Neural Information Processing (ICONIP2020, CORE ranking-A)
Yuanzhe Zhang, Ling Luo, Jianjia Zhang et al.
This work investigated the next basket recommendation task and proposed the first end-to-end correlation-aware model to predict the next basket considering intra-basket correlations using graph attention networks. By estimating and aggregating the intra-basket correlations using the attention layer of the self-attention model, the recommendation can be conducted at the basket level, instead of at the item level.
FCP Filter: A Dynamic Clustering-Prediction Framework for Customer Behavior at the 24th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD 2020, CORE ranking-A)
Yuanzhe Zhang, Ling Luo, Yang Wang, Zhiyong Wang.
Customer purchase behaviour prediction is often limited by the randomness of individual historic transaction data. In the meanwhile, Fragmentation and Coagulation Process (FCP), as a SOTA stochastic customer partition model, does not support forecasting purchase behaviour as such a data-driven method requires transaction observations to conduct clustering. To tackle this challenge, this work proposed FCP filter, a clustering-prediction framework based on FCP. FCP filter utilizes FCP to cluster customers into groups by their temporal interests to filter the individual random noise. A predictor is then built on grouped data and the predicted results are fed to FCP as the prior knowledge at the next time step. We demonstrated that FCP filter can discover a flexible number of latent groups and provide accurate predictions for dynamic purchase behaviour.