Source-linked AI summary
Analysis of Cryptocurrency Transactions from a Network Perspective: An Overview
Jiajing Wu, Jieli Liu, Yijing Zhao, Zibin Zheng
TL;DR
Cryptocurrency transaction records are publicly accessible, but their analysis requires a systematic account of heterogeneous network representations, descriptive studies, and detection tasks. This survey reviews the literature across network modeling, profiling, and network-based detection, finding a broad body of methods and results while identifying limitations in transaction tracing and pattern-based approaches. It concludes with research directions and resources for future work.
Problem
Public cryptocurrency transaction records enable analysis, but existing work spans heterogeneous network representations, profiling studies, and detection tasks that require systematic synthesis.
Method
The survey organizes state-of-the-art literature into network modeling, network profiling, and network-based detection, covering methods, findings, datasets, and future directions.
Results
The reviewed literature covers heterogeneous graph models, descriptive analyses of network properties and evolution, and detection tasks including entity recognition, illicit activity detection, and transaction tracing.
Takeaways & Limitations
Cryptocurrency transaction analysis can be studied through complementary modeling, profiling, and detection perspectives, with entity recognition often supporting downstream investigative tasks.
Takeaways & Limitations
Transaction-tracing methods are mostly simple and heuristic, and except for cross-ledger tracing they mainly suit Bitcoin-like transaction-centered platforms.
Abstract
from arXiv · showhide
As one of the most important and famous applications of blockchain technology, cryptocurrency has attracted extensive attention recently. Empowered by blockchain technology, all the transaction records of cryptocurrencies are irreversible and recorded in the blocks. These transaction records containing rich information and complete traces of financial activities are publicly accessible, thus providing researchers with unprecedented opportunities for data mining and knowledge discovery in this area. Networks are a general language for describing interacting systems in the real world, and a considerable part of existing work on cryptocurrency transactions is studied from a network perspective. This survey aims to analyze and summarize the existing literature on analyzing and understanding cryptocurrency transactions from a network perspective. Aiming to provide a systematic guideline for researchers and engineers, we present the background information of cryptocurrency transaction network analysis and review existing research in terms of three aspects, i.e., network modeling, network profiling, and network-based detection. For each aspect, we introduce the research issues, summarize the methods, and discuss the results and findings given in the literature. Furthermore, we present the main challenges and several future directions in this area.
1. Introduction
This survey examines cryptocurrency transaction mining from a network perspective, organizing prior work into network modeling, profiling, and network-based detection. It also summarizes research findings, future directions, and resources for researchers.
- Scope: The survey reviews cryptocurrency transaction-network research published primarily from January 2009 to May 2020.It also includes selected book chapters, theses, arXiv papers, and technical reports.
- Network modeling: Network modeling constructs graphs for transaction information and categorizes methods by node and edge semantics.The survey focuses on how different graph representations support various analytical tasks.
- Network profiling: Network profiling describes transaction networks through network properties, network evolution, and market effects.These studies provide empirical or analytical overviews of cryptocurrency transaction networks.
- Network-based detection: Network-based detection covers entity recognition, transaction pattern recognition, illicit activity detection, and transaction tracing.Entity recognition commonly supports downstream tasks, while tracing follows money flows associated with detected illicit activities.
- Contributions: The survey provides a structured overview of network construction, profiling, and detection, summarizes important findings, and proposes future research directions.It also compiles benchmark data sources and guidance for new researchers.
2. Background
The background introduces cryptocurrency development, transaction models, data acquisition, and labeling resources. It contrasts Bitcoin's transaction-centered model with Ethereum's account-centered model and describes practical sources for network analysis.
- Cryptocurrency: Cryptocurrencies developed from earlier digital-currency proposals to Bitcoin and later alternative cryptocurrencies such as Ethereum, Litecoin, and Ripple.The passage describes Bitcoin as the first decentralized cryptocurrency and Ethereum as a smart-contract platform.
- Transaction process: Blockchain transactions are broadcast through a peer-to-peer network and recorded in blocks, while transaction models are generally transaction-centered or account-centered.Bitcoin and Ethereum exemplify these two models, respectively.
- Bitcoin: Bitcoin uses addresses and multi-input, multi-output transactions built from unspent transaction outputs.A user can possess multiple addresses, supporting pseudonymity.
- Ethereum: Ethereum uses externally owned accounts and contract accounts, with contract accounts maintaining executable bytecode and state information.Externally owned accounts can create and invoke smart contracts.
- Data sources: Researchers can obtain raw blockchain data by synchronizing blocks with clients, parsing binary data, and using Ethereum trace data for internal transaction records.The survey also identifies label services and processed datasets for specialized mining tasks.
3. Network Modeling
Cryptocurrency transaction networks require modeling choices tailored to heterogeneous blockchain data and platform-specific transaction models. Existing approaches represent money flows, ownership, contract operations, account creation, or voting through networks whose node and edge semantics depend on the task.
- Modeling challenges: Network modeling converts cryptocurrency transaction data into network-structured data before analysis.The modeling process must account for heterogeneous blockchain data from multiple sources.
- Modeling challenges: Heterogeneous objects and platform-specific data structures make node and link definitions difficult to construct accurately.Blockchain platforms differ in their data structures, so modeling methods are cryptocurrency-oriented.
- UTXO-based systems: UTXO-based systems support transaction networks with transactions as nodes and user networks with entities as nodes connected by money flows.User networks rely on entity recognition to cluster addresses, but ownership uncertainty limits their accuracy.
- Account-centered systems: Account-centered systems use distinct graphs for money flow, contract creation, and contract invocation, with edge semantics determined by the represented operation.Ethereum’s MFG, CCG, and CIG include externally owned accounts and contract accounts as nodes.
- Account-centered systems: EOS modeling additionally represents account creation and account voting because these operations are stored as transaction actions.These graphs extend account-centered modeling to operations associated with delegated proof-of-stake.
- Task-specific modeling: Network construction should preserve the transaction information required by a specific analytical task, including temporal information when needed.Ethereum transaction records have been modeled as temporal weighted multidigraphs.
4. Network Profiling
Network profiling characterizes cryptocurrency transaction networks through structural properties and their evolution over time. Studies report heterogeneous degree patterns, varying connectivity and clustering, changing communities, and distinct development phases across cryptocurrencies and platforms.
- Network size and degree: Bitcoin transaction networks grow faster than linearly, and their densification follows a power-law rule.Ethereum studies also compare node and edge growth across multiple network representations.
- Network size and degree: Cryptocurrency transaction networks often exhibit heterogeneous degree distributions that follow power-law patterns, with the fitted parameter α eventually reaching a steady state.The reported steady-state behavior follows fluctuations in α across five cryptocurrencies.
- Path and clustering structure: Bitcoin user networks across countries have similar average shortest paths, while diameter changes have been used to study anonymity and other structural factors.Shortest-path analyses also assess small-world behavior in Bitcoin and Ethereum.
- Path and clustering structure: Bitcoin and Ethereum show relatively high clustering and small-world properties, whereas Ripple and Namecoin transaction networks have relatively low clustering coefficients.The clustering coefficient measures closure among paths of length two.
- Centrality and connectivity: Centrality measures identify important nodes: exchanges act as major hubs, and PageRank-selected ERC20 traders transact frequently with one another.The reviewed measures include degree, eigenvector, Katz, PageRank, betweenness, and closeness centrality.
- Centrality and connectivity: Connected-component analysis shows that Bitcoin’s component count rose before 2011 and later shrank as exchanges promoted bitcoin circulation.Most Bitcoin addresses were reported within the largest connected component.
- Communities and motifs: Community studies find dynamic user communities in Ripple, while Bitcoin transaction networks tend toward modular, mostly tree-like community structures.Ripple communities were observed to form through connections to gateways.
- Network evolution: Temporal analyses distinguish major development phases, including Bitcoin’s initial and trading phases and Ethereum’s slow-start, outbreak, and fever-abatement phases.Evolution studies also relate network dynamics to events and exchange-rate changes, including an Ether price spike near the 5,000,000th block.
5. Network-based Detection
Network-based detection uses transaction data to identify entities, recognize transaction patterns, detect illicit activity, and trace transactions. The reviewed methods exploit transaction properties, behavioral similarities, visualization, network embeddings, and taint analysis, but remain limited by false positives, platform specificity, delayed detection, and heuristic tracing.
- Entity Recognition: Entity recognition clusters pseudonymous addresses belonging to the same user, forming a basis for downstream detection tasks.Methods include transaction-property heuristics, behavior-based clustering, and off-chain information for de-anonymization.
- Entity Recognition: Transaction-property heuristics use multi-input, change-address, and platform-specific rules to associate addresses with common entities.The multi-input heuristic assumes transaction inputs share ownership, while change-address methods link outputs receiving change to input owners.
- Entity Recognition: Heuristic entity recognition can produce underestimation or overestimation errors, while behavioral methods are relatively coarse-grained.Combining multi-input heuristics with community detection can increase recall at the cost of precision.
- Transaction Pattern Recognition: Transaction pattern recognition summarizes common and special-address behaviors, supporting illicit-activity detection through visualization and network-based classifiers.Studies identify patterns such as money laundering and denial-of-service attacks, while network embeddings have been used to classify money-laundering transactions.
- Illicit Activity Detection: Illicit-activity detection relies on public, irreversible transaction records but most methods report crimes only after they occur, lacking early warning and interception.Money-laundering detection includes feature analysis, network embeddings such as DeepWalk and node2vec, and purity-based analysis of bitcoin mixing.
- Transaction Tracing: Transaction-tracing methods are mostly simple and heuristic, and generally target Bitcoin-like transaction-centered platforms rather than more complex blockchain settings.The survey identifies risk propagation under more complex factors and scenarios as a future direction.
6. Discussion & Future Research Directions
The survey organizes cryptocurrency transaction network analysis around modeling, profiling, and detection, then identifies challenges involving heterogeneous, incomplete, rapidly growing data and privacy-preserving techniques. Future directions emphasize compatible modeling, information completion, dynamic online analysis, and feasible auditing and tracing.
- Research organization: Cryptocurrency transaction network analysis is organized into network modeling, network profiling, and network-based detection.Modeling abstracts transaction data into networks; profiling characterizes network properties, evolution, and market effects; detection covers entity recognition, transaction pattern recognition, illicit activity detection, and transaction tracing.
- Challenges: Existing detection methods can be bypassed by privacy-enhancing techniques, and rapidly growing transaction data hinders real-time illicit-activity detection.These limitations affect transaction-property methods, transaction tracing, and graph-based anomaly detection, indicating a need for stronger and online-capable approaches.
- Future research directions: Different blockchain transaction models and structures require compatible network modeling methods across cryptocurrencies and analytical tasks.Bitcoin is transaction-centered, Ethereum is account-centered and includes contract accounts, and EOS transactions can contain multiple actions.
- Future research directions: Off-chain solutions such as Lightning Network can omit internal transaction links, motivating link prediction and other methods to complement incomplete network information.Lightning Network records channel opening and closing while omitting transactions within the channel; predicted links can restore partially missing structures.
- Future research directions: Hundreds of millions of Bitcoin and Ethereum transaction records and continually appearing nodes and edges create challenges for dynamic analysis and online learning.The rapid growth of transaction data makes real-time deployment of illicit-activity detection difficult and motivates online algorithms.
- Future research directions: Privacy-enhancing techniques make transaction auditing and tracing more difficult, while their abuse can accompany scams and money laundering.Mixing services, ring signatures, and non-interactive zero-knowledge proofs can hinder tracing; network methods are proposed for auditing, tracing, and interception.
7. Conclusion
The paper reviews cryptocurrency transaction-network research and organizes it into modeling, profiling, and network-based detection. It presents these networks as unusually valuable for graph-mining research because they are publicly accessible, rapidly evolving, and rich in temporal, value, and label information, while positioning the field as a foundation for future methods and findings.
- Conclusion: The survey reviews state-of-the-art cryptocurrency transaction-network literature through three mutually supportive steps: modeling, profiling, and network-based detection.It also introduces key concepts in cryptocurrency transactions and categorizes existing techniques and results within this framework.
- Research significance: Publicly accessible cryptocurrency transaction networks are among the largest real-world networks available for research.They also contain rapidly evolving structures and abundant temporal, value, and label information.
- Future outlook: The field offers opportunities to advance complex network theory and graph-mining techniques through large, dynamic, information-rich transaction networks.The paper describes cryptocurrency transaction networks as a promising interdisciplinary area for new findings, methods, and innovations.