Fraudulent activities on blockchain networks threaten the integrity and reliability of decentralized finance ecosystems. Accurately identifying malicious nodes such as phishing or ransomware addresses, within large-scale blockchain transaction graphs remains a critical challenge due to their dynamic, sparse, and continuously evolving topologies. Transfer learning offers a powerful paradigm for fraud detection because many fraudulent schemes, including ransomware and phishing, are often orchestrated by overlapping actor groups that share behavioral and structural patterns across networks. Leveraging these shared representations enables knowledge transfer from previously observed fraud types to emerging ones. However, the complex and multi-modal nature of digital financial systems introduces substantial challenges for graph-based transfer learning. Fraudulent activities are shaped by diverse modalities including graph structure, transaction sequences, temporal price dynamics, and textual metadata, while distributional shifts frequently occur across time and platforms. Existing graph transfer learning methods struggle to model such multi-modal dependencies and to align divergent feature distributions. To tackle these challenges, we develop a Multi-mOdal Enhanced Graph Transfer Learning (MOE-GTL) framework which incorporates graph, temporal, and textual modalities for fraudulent node detection. We further introduce Temporal-aware Maximum Mean Discrepancy (TMMD), a regularization mechanism that explicitly aligns multi-modal feature distributions between source and target graphs over time. Extensive experiments reveal that our MOE-GTL model notably improves the accuracy of fraudulent node classifications on Ethereum and Solana transaction graphs.

Fraudulent activity on blockchain networks poses significant risks to the integrity and trust of decentralized finance ecosystems. The timely and accurate detection of fraud nodes such as phishing addresses within large-scale Ethereum transaction networks remains a major challenge due to their dynamic, sparse, and evolving structures. While methods like graph deep learning (e.g., graph neural networks) have been extensively explored, they are not inherently designed to capture higherorder interactions and textual information embedded within graph data. Motivated by the urgent need for advanced and robust fraud detection techniques, we introduce a novel graph prompting method named Large Language Model-Simplicial Complex (LLM-SC) based graph prompting framework that leverages LLM-based multi-agent collaboration system, LLMbased financial news prompt function, and simplicial neural networks to capture both the structural and contextual dimensions of blockchain activity. The empirical studies demonstrate the effectiveness of our approach, and these results provide a new tool for blockchain analytics platforms and regulatory authorities, enabling earlier and more accurate identification of fraudulent behavior and ultimately supporting safer and more resilient digital financial systems. The code is available at https://github.com/y13564/LLM-SC.

▪Challenges and Motivation ▪Silicon Features ▪Vision Applications