BlockScan: Detecting Anomalies in Blockchain Transactions

TL;DR

BlockScan introduces a specialized Transformer model with novel multi-modal tokenization and pretraining techniques for effective anomaly detection in blockchain transactions, outperforming existing methods on Ethereum and Solana.

Contribution

The paper presents a customized Transformer architecture with unique tokenization and pretraining strategies specifically designed for blockchain transaction data, setting a new benchmark in anomaly detection.

Findings

High detection accuracy on Ethereum transactions

Successful anomaly detection on Solana transactions

Outperforms existing methods with low false positive rate

Abstract

We propose BlockScan, a customized Transformer for anomaly detection in blockchain transactions. Unlike existing methods that rely on rule-based systems or directly apply off-the-shelf large language models (LLMs), BlockScan introduces a series of customized designs to effectively model the unique data structure of blockchain transactions. First, a blockchain transaction is multi-modal, containing blockchain-specific tokens, texts, and numbers. We design a novel modularized tokenizer to handle these multi-modal inputs, balancing the information across different modalities. Second, we design a customized masked language modeling mechanism for pretraining the Transformer architecture, incorporating RoPE embedding and FlashAttention for handling longer sequences. Finally, we design a novel anomaly detection method based on the model outputs. We further provide theoretical analysis for the detection method of our system. Extensive evaluations on Ethereum and Solana transactions demonstrate BlockScan's exceptional capability in anomaly detection while maintaining a low false positive rate. Remarkably, BlockScan is the only method that successfully detects anomalous transactions on Solana with high accuracy, whereas all other approaches achieved very low or zero detection recall scores. This work sets a new benchmark for applying Transformer-based approaches in blockchain data analysis.