FlowShield：基于交易语义解析与资金流追踪的加密货币反洗钱
FlowShield: cryptocurrency anti-money laundering with transaction semantics parsing and fund flow tracking
浏览论文内容
中文总结 AI 辅助
FlowShield是一种加密货币反洗钱框架，通过解析交易语义、重构资金流子图并结合LLM与GCN实现洗钱检测，在BybitML等数据集上平均F1达98.0%，还可生成可读报告辅助调查。
中文摘要 AI 辅助
加密货币反洗钱（Crypto AML）正日益面临复杂洗钱行为的挑战，此类行为通过多样化语义及跨多条区块链快速拆分被盗资产。现有Crypto AML方法常简化交易语义、依赖以拓扑为中心的信号，或输出孤立的检测标签。本文提出FlowShield——一种面向交易级洗钱检测及面向调查人员的报告生成的Crypto AML框架。FlowShield首先从可观测关系中恢复行为级语义，明确洗钱意图；为追踪价值来源与再分配，FlowShield从三个互补视角重构资金流子图；随后采用文本-结构融合机制，实现大语言模型（LLM）编码的语义与图卷积网络（GCN）编码的结构之间的交互。除检测外，FlowShield还生成可读的可疑活动报告（SARs），为调查人员提供简洁摘要与可解释的风险信号。为解决多链检测中的数据稀缺问题，我们构建并开源了首个公开多链洗钱数据集BybitML。我们在BybitML及两个公开洗钱数据集上评估FlowShield，实验结果显示FlowShield取得最佳整体性能，平均F1分数达98.0%；进一步的行为与SAR分析表明，FlowShield可揭示多样化洗钱策略，并生成可读报告以调查复杂多跳资金流。
英文摘要
Cryptocurrency anti-money laundering (Crypto AML) is increasingly challenged by sophisticated laundering behaviors that rapidly fragment stolen assets through diverse semantics and across multiple blockchains. Existing Crypto AML methods often simplify transaction semantics, rely on topology-centric signals, or output isolated detection labels. In this paper, we present \textsc{FlowShield}, a Crypto AML framework for transaction-level laundering detection and investigator-facing report generation. \textsc{FlowShield} first recovers behavior-level semantics from observable relations, making laundering intents explicit. To trace value provenance and redistribution, \textsc{FlowShield} reconstructs fund-flow subgraphs from three complementary perspectives. It then employs a text--structure fusion mechanism, enabling the interplay between large language model (LLM)-encoded semantics and flow texts with graph convolutional network (GCN)-encoded structure. Beyond mere detection, \textsc{FlowShield} further generates readable suspicious activity reports (SARs), offering investigators concise summaries and explainable red flags. To address the data scarcity in multi-chain detection, we construct and open-source \textit{BybitML}, the first public multi-chain laundering dataset. We evaluate \textsc{FlowShield} on \textit{BybitML} and two public laundering datasets and experimental results demonstrate that \textsc{FlowShield} achieves the best overall performance, with an average F1 score of 98.0\%. Further behavior and SAR analyses demonstrate that \textsc{FlowShield} can reveal diverse laundering strategies and produce readable reports for investigating complex multi-hop fund flows.
