Graph Inference for Tracking Obfuscated Blockchain Transactions

This direction is merged into another direction: blockchain pseudonymity. Its last version stays readable and citable here.

Version 1 of 112 papers16 conceptsfrozen 2026-10-04

Public blockchains like Bitcoin and Ethereum are often described as transparent ledgers, but for forensic investigators, this transparency is increasingly an illusion. Criminal actors have moved beyond simple direct transfers, adopting privacy-enhancing technologies (PETs) that intentionally "break" the transaction graph.

To counter this, researchers are shifting from simple link-tracing to sophisticated graph inference. Instead of following a single path of money, these new methods analyze the topological "shape," temporal evolution, and structural patterns of transactions to unmask illicit activity.

The Problem of Broken Edges

Traditional blockchain forensics relies on "taint analysis"—the idea that if Address A sends money to Address B, the funds can be traced. However, as Seunghyeon Lee and colleagues demonstrate in their study of Dark Web cryptocurrency abuses, malicious actors now use stealth addresses, ring signatures, and zero-knowledge proofs to obscure these edges.

These technologies create "dark graphs." As Jiajing Wu explains in a comprehensive overview, mixers and tumblers act as black boxes: illicit funds go in, are shuffled with legitimate funds, and come out at a different address. This severs the observable chain of custody, making standard heuristic analysis ineffective.

Inferring Identity from Graph Topology

If the direct links are broken, how do we track the money? The key insight is that while individual transactions can be hidden, the behavioral footprint of a criminal operation often remains visible in the graph's topology.

Cuneyt Gurcan Akcora and Yulia Gel pioneered the use of Topological Data Analysis (TDA) for this purpose. In their BitcoinHeist framework, they don't just look at who sent money to whom; they map the "shape" of transaction patterns. Ransomware families, for instance, exhibit specific patterns of splitting and merging funds that are distinct from legitimate commercial activity. By focusing on these persistent topological features, they can predict ransomware transactions even when the specific addresses are new.

Combining Global Snapshots with Local Evolution

Modern de-anonymization frameworks now treat the blockchain as a dynamic, evolving system rather than a static map. Shuyi Miao and Yang Liu developed DBG4ETH, which uses a "double graph" architecture.

This approach works by looking at two things simultaneously:

  1. Global Topology: A snapshot of how an account relates to the entire Ethereum ecosystem.
  2. Local Evolution: How that account’s specific transaction patterns change over time.

By integrating these views, the system can identify the real-world entities behind anonymous addresses, even when those entities use mixers to disrupt their transaction history. Similarly, Zongwei Li and Xiaoqi Li developed ScamSweeper, which uses temporal sampling and Transformer architectures to distinguish "mimic" scam accounts from legitimate services by analyzing how their transaction networks evolve.

The rise of Decentralized Finance (DeFi) has added another layer of obfuscation. A single "transaction" might now involve complex token swaps across multiple liquidity pools. To solve this, Zhiying Wu and Jieli Liu introduced TRacer, a tool that utilizes a modified version of the Personalized PageRank algorithm.

Instead of ranking web pages, TRacer ranks blockchain nodes based on their likelihood of being "cash-out" points for stolen funds. By understanding the logic of DeFi protocols, TRacer can propagate "taint" through complex smart contract interactions that would baffle traditional tracing tools. This allows investigators to produce small, human-auditable subgraphs that pinpoint where illicit assets eventually exit the ecosystem.

From Individual Actors to Campaigns

Finally, research is moving toward identifying entire "campaigns" rather than isolated accounts. Pengcheng Xia and Kailong Wang demonstrated that by combining web crawling with blockchain analysis, it is possible to group thousands of seemingly unrelated addresses into distinct criminal operations. Their work reveals that the Dark Web's financial ecosystem is dominated by a small number of "long-living" actors who manage sophisticated, multi-year campaigns.

This shift in perspective—from tracing coins to inferring the structural and temporal logic of the actors—represents the new frontier in blockchain forensics. As Jiqiang Tang notes, while regulatory challenges remain in a "centerless" financial world, these multi-dimensional monitoring frameworks are beginning to close the gap between anonymity and accountability.

Papers in this version

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_graph_inference_for_tracking_obfuscated__549f587b,
  title = {Graph Inference for Tracking Obfuscated Blockchain Transactions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/graph-inference-for-tracking-obfuscated-blockchain-transactions/rdr_549f587b09fb493c949045208c38f723},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_graph_inference_for_tracking_obfuscated__549f587b_v1,
  title = {Graph Inference for Tracking Obfuscated Blockchain Transactions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/graph-inference-for-tracking-obfuscated-blockchain-transactions/rdr_549f587b09fb493c949045208c38f723/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}