Optimizing Blockchain Analysis: Tackling Temporality and Scalability with an Incremental Approach with Metropolis-Hastings Random Walks

TL;DR

This work tackles temporality and scalability in blockchain transaction network analysis by combining an Incremental Unbiased Update with Metropolis-Hastings random walks to produce efficient, informative node embeddings. The approach enables updating embeddings without re-training the entire graph while guiding walk exploration toward high-utility, $h$-hop regions, reducing the number of walks needed. Empirical results across multiple datasets show that MH-based embeddings achieve competitive or superior node classification performance with fewer walks, and the Unbiased Update closely matches From-Scratch embeddings, demonstrating practical scalability for growing blockchain graphs. The methodology offers a scalable tool for transaction monitoring, fraud detection, and address-type classification in real-world, rapidly expanding blockchain networks.

Abstract

Blockchain technology, with implications in the financial domain, offers data in the form of large-scale transaction networks. Analyzing transaction networks facilitates fraud detection, market analysis, and supports government regulation. Despite many graph representation learning methods for transaction network analysis, we pinpoint two salient limitations that merit more investigation. Existing methods predominantly focus on the snapshots of transaction networks, sidelining the evolving nature of blockchain transaction networks. Existing methodologies may not sufficiently emphasize efficient, incremental learning capabilities, which are essential for addressing the scalability challenges in ever-expanding large-scale transaction networks. To address these challenges, we employed an incremental approach for random walk-based node representation learning in transaction networks. Further, we proposed a Metropolis-Hastings-based random walk mechanism for improved efficiency. The empirical evaluation conducted on blockchain transaction datasets reveals comparable performance in node classification tasks while reducing computational overhead. Potential applications include transaction network monitoring, the efficient classification of blockchain addresses for fraud detection or the identification of specialized address types within the network.

Figures (3)

• Figure 1: Overview of the address (node) classification pipeline in a blockchain transaction network. Starting from raw transactions in blockchain blocks transformed into a transaction graph. Random walk method is performed to generate node embeddings. Our method (illustrated in the rectangular container) is towards addressing the challenges of incremental learning and scalable solution of optimizing walk efficiency.
• Figure 2: Illustration of the Metropolis-Hastings Random Walk (Leap-walk) on a transaction graph, where nodes represent addresses and edges represent token transfers. The highlighted dotted curved pathway exemplifies that leap-walk exploration enables sampling beyond immediate neighbors.
• Figure 3: Comparison of mean defacto walk lengths across methods for a fixed walk length of 5. MH methods consistently achieved a higher mean walk length compared to other methods, emphasizing its ability to traverse more unique nodes within the same walk length constraints.