Paper: arXiv 2503.01686

Authors: Honglin Fu, Yebo Feng, Cong Wu, Jiahua Xu

Abstract

Masterminds are entities organizing, coordinating, and orchestrating cryptocurrency pump-and-dump schemes, a form of trade-based manipulation undermining market integrity and causing financial losses for unwitting investors. Previous research detects pump-and-dump activities in the market, predicts the target cryptocurrency, and examines investors and \ac{“osn”} entities. However, these solutions do not address the root cause of the problem. There is a critical gap in identifying and tracing the masterminds involved in these schemes. In this research, we develop a detection system \textsc{“Perseus”}, which collects real-time data from the \acs{“osn”} and cryptocurrency markets. \textsc{“Perseus”} then constructs temporal attributed graphs that preserve the direction of information diffusion and the structure of the community while leveraging \ac{“gnn”} to identify the masterminds behind pump-and-dump activities. Our design of \textsc{“Perseus”} leads to higher F1 scores and precision than the \ac{“sota”} fraud detection method, achieving fast training and inferring speeds. Deployed in the real world from February 16 to October 9 2024, \textsc{“Perseus”} successfully detects \(438\) masterminds who are efficient in the pump-and-dump information diffusion networks. \textsc{“Perseus”} provides regulators with an explanation of the risks of masterminds and oversight capabilities to mitigate the pump-and-dump schemes of cryptocurrency.

Complexity vs Empirical Score

  • Math Complexity: 2.5/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper applies Graph Neural Networks (GNNs) and temporal graph modeling, which involves moderate mathematical complexity, but the focus is overwhelmingly on system implementation, real-world deployment, and extensive empirical validation. It features detailed data collection, large-scale real-time deployment, and high-performance metrics (F1, precision), indicating a practical, backtest-ready system.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Identify & Trace Masterminds<br>Behind Crypto Pump-and-Dump Schemes"] --> B["Data Collection & Preprocessing<br>Real-time OSN & Market Data"]
  B --> C["Temporal Graph Construction<br>Attributed Graphs preserving<br>Info Diffusion & Community Structure"]
  C --> D["GNN-based Detection<br>Leveraging Graph Neural Networks<br>to Identify Masterminds"]
  D --> E["Real-World Deployment<br>Feb 16 - Oct 9, 2024"]
  E --> F{"Key Outcomes"}
  F --> F1["Detected 438 Masterminds<br>Efficient in Diffusion Networks"]
  F --> F2["Higher F1 & Precision vs SOTA<br>Fast Training & Inference"]
  F --> F3["Risk Explanation & Oversight<br>Tools for Regulators"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes PERSEUS, a detection system that uses real-time OSN and cryptocurrency market data with graph neural networks to identify mastermind entities behind pump-and-dump schemes.

Source sentence
In this research, we develop a detection system PERSEUS, which collects real-time data from the OSN and cryptocurrency markets.

Assumes 2

  • At least one node in each temporal attributed graph is a mastermind organizing crowd‑pump events. model
    Source sentence
    We assume that at least one node, vm ∈V (Dx τ ), is the mastermind organizing crowd-pump events.
  • The maximum return is measured within a three‑day window after the announcement because crowd‑pump effects typically do not last longer. data
    Source sentence
    The three‑day window is chosen because crowd‑pump effects typically do not last longer.

Data

  • Binance cryptocurrency market data · Binance market-prices public
    February 16 to October 9 2024 · cryptocurrencies
    Source sentence
    PERSEUS successfully detects 438 masterminds who are efficient in the pump-and-dump information diffusion networks.
  • Telegram OSN messages · Telegram text-news public
    February 16 to October 9 2024 · crowd‑pump channels
    Source sentence
    OSNs are increasingly exploited by spreaders to enhance the effects of crowd‑pumps. Specifically, on platforms like Telegram, spreaders establish channels where only they can talk to coordinate investors.

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • PERSEUS achieves higher detection performance than the state‑of‑the‑art fraud detection method, with weighted diffusion GraphSAGE attaining the highest ROC AUC of 0.93. predictive-accuracy
    Source sentence
    Weighted diffusion GraphSAGE achieving the highest ROC AUC (0.93 vs. 0.92 for directed), and weighted diffusion GAT (0.87) surpassing its directed counterpart (0.84).
  • GraphSAGE outperforms GAT in precision and F1 score across threshold ranges, with precision peaking around 0.55 and F1 around 0.5. predictive-accuracy
    Source sentence
    GraphSAGE outperforms GAT in precision within the 0.50–0.6 threshold range, with similar performance elsewhere. The F1 score
  • Training and inference are fast: GraphSAGE models train each epoch in approximately 0.13 to 0.18 seconds, while GAT models take 0.23 to 0.36 seconds. performance
    Source sentence
    GraphSAGE models take approximately 0.13 to 0.18 seconds to train each epoch, while the GAT models take more time, around 0.23 to 0.36 seconds.
  • The system detected 438 mastermind entities in real‑world deployment. other
    Source sentence
    PERSEUS successfully detects 438 masterminds who are efficient in the pump-and-dump information diffusion networks.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 46 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (100)

  1. Market Abuse Regulation 1.6 — link
  2. A CFTC Primer on Virtual Currencies (2017) — link
  3. A Novel Text Mining Approach to Securities and Financial Fraud Detection of Case Suspects
  4. An examination of the cryptocurrency pump-and-dump ecosystem (2021) — link
  5. Analyzing Target-Based Cryptocurrency Pump and Dump Schemes (2021) — link
  6. Anomaly Detection with Graph Convolutional Networks for Insider Threat and Fraud Detection (2019) — link
  7. Charting the Landscape of Online Cryptocurrency Manipulation (2020) — link
  8. Commodity Exchange Act — link
and 92 more
  1. Crypto Pump and Dump Detection via Deep Learning Techniques (2022) — link
  2. cryptocompare _ volume (2017) — link
  3. Cryptocurrency Pump and Dump Schemes: Quantification and Detection (2019) — link
  4. Cryptocurrency Pump-and-Dump Schemes (2022) — link
  5. DANI: A Fast Diffusion Aware Network Inference Algorithm (2017)
  6. Data Collection and Analysis Phase: Part III (2012) — link
  7. Deep Anomaly Detection on Attributed Networks (2019) — link
  8. Detecting ‘Pump & dump schemes’ on cryptocurrency market using an improved apriori algorithm (2019) — link
  9. Detecting cryptocurrency pump-and-dump frauds using market and social signals (2021) — link
  10. Diga: Guided Diffusion Model for Graph Recovery in Anti-Money Laundering (2023) — link
  11. Digital civilisation: manifesto for a trustworthy, well regulated world. For the SSC Blockchain initiative and the digital civilisation conference — link
  12. Final rules and amendments to joint industry plans (2005) — link
  13. Financial Conduct Authority Handbook — link
  14. Graph Attention Networks (2017)
  15. Identifying and Analyzing Cryptocurrency Manipulations in Social Media (2021) — link
  16. Inductive Representation Learning on Large Graphs (2017)
  17. Influencers, Amplifiers, and Icons: A Systematic Approach to Understanding the Roles of Islamophobic Actors on Twitter (2022) — link
  18. Introduction to Social Network Methods (2005) — link
  19. Is Bitcoin Really Untethered? (2020) — link
  20. Limelight: Real-Time Detection of Pump-and-Dump Events on Cryptocurrency Exchanges Using Deep Learning (2019) — link
  21. Profitability and herding of trade-based pump-and-dump manipulation (2024) — link
  22. Profitability of cryptocurrency Pump and Dump schemes (2021) — link
  23. Pump and Dumps in the Bitcoin Era: Real Time Detection of Cryptocurrency Market Manipulations (2020) — link
  24. Regulation (EU) 2023/1114 of the European Parliament and of the Council of 31 May 2023 on markets in crypto-assets, and amending Regulations (EU) No 1093/2010 and (EU) No 1095/2010 and Directives 2013/36/EU and (EU) 2019/1937 (Text with EEA relevance) (2023) — link
  25. SEC Charges Kim Kardashian for Unlawfully Touting Crypto Security (2022) — link
  26. SEC Files Charges in a Crypto Asset Pump-And-Dump Scheme (2022) — link
  27. Selective Disclosure and Insider Trading (2000) — link
  28. Semi-Supervised Classification with Graph Convolutional Networks (2016)
  29. Sequence-based target coin prediction for cryptocurrency pump-and-dump (2022) — link
  30. Text of H.R. 4763, The Financial Innovation and Technology for the 21st Century Act (2024)
  31. The Anatomy of a Cryptocurrency Pump-and-Dump Scheme (2019) — link
  32. The danger of cryptocurrency pump-and-dumps: Analysing the development of research & regulation (2021) — link
  33. The Doge of Wall Street: Analysis and Detection of Pump and Dump Cryptocurrency Manipulations (2022) — link
  34. To the moon: defining and detecting cryptocurrency pump-and-dumps (2018) — link
  35. TTAGN: Temporal Transaction Aggregation Graph Network for Ethereum Phishing Scams Detection (2022) — link
  36. User profiling in an ego network (2014) — link
  37. Web 3.0 Tokenization and Decentralized Finance (DeFi) (2022) — link
  38. Web 3.0 Tokenization and Decentralized Finance (DeFi) (2022) — link
  39. Why Are Cryptocurrencies Vulnerable To Pump-and-Dump Schemes in Europe?? (2022) — link
  40. 2017 SSCI proceedings
  41. A Survey on Deep Learning for Named Entity Recognition (2022) — link
  42. Analyzing the Bitcoin Ponzi scheme ecosystem (2019) — link
  43. Breaking the Anonymity of Ethereum Mixing Services Using Graph Feature Learning (2024) — link
  44. Cascading Machine Learning to Attack Bitcoin Anonymity (2019) — link
  45. Cashtag piggybacking: Uncovering spam and bot activity in stock microblogs on twitter (2019) — link
  46. Characterizing speed and scale of cryptocurrency discussion spread on reddit (2019) — link
  47. Cloudburst (2023) — link
  48. Crypto-influencers (2024) — link
  49. Cryptocurrency price drivers: Wavelet coherence analysis revisited (2018) — link
  50. Data mining for detecting Bitcoin Ponzi schemes (2018) — link
  51. Detecting malicious accounts in permissionless blockchains using temporal graph properties (2021) — link
  52. Digital Economy: The Economics of the Digital Economy (2023) — link
  53. Digital Tokens and Platform Building (2018) — link
  54. Dynamic Topic Modelling for Cryptocurrency Community Forums (2017) — link
  55. Edge Directionality Improves Learning on Heterophilic Graphs (2023)
  56. Fishing for Fraudsters: Uncovering Ethereum Phishing Gangs With Blockchain Data (2024) — link
  57. Graph-Based Covert Transaction Detection and Protection in Blockchain (2024) — link
  58. Identifying and Analyzing Cryptocurrency Manipulations in Social Media (2019) — link
  59. Illicit Social Accounts? Anti-Money Laundering for Transactional Blockchains (2025) — link
  60. Inductive Representation Learning on Temporal Graphs (2020)
  61. Joint Inference of Diffusion and Structure in Partially Observed Social Networks Using Coupled Matrix Factorization (2023) — link
  62. Machine Learning-Based Detection of Pump-and-Dump Schemes in Real-Time (2024)
  63. Market Manipulation and Suspicious Stock Recommendations on Social Media (2017) — link
  64. MT 2 AD: multi-layer temporal transaction anomaly detection in ethereum networks with GNN (2024) — link
  65. On Inductive–Transductive Learning With Graph Neural Networks (2022) — link
  66. Pairwise-interactions-based Bayesian Inference of Network Structure from Information Cascades (2023) — link
  67. Pairwise-interactions-based Bayesian Inference of Network Structure from Information Cascades (2023) — link
  68. Predicting cryptocurrency price bubbles using social media data and epidemic modelling (2018) — link
  69. Predicting Fluctuations in Cryptocurrency Transactions Based on User Comments and Replies (2016) — link
  70. Regulation (EU) 2023/1114 of the European Parliament and of the Council of 31 May 2023 on markets in crypto-assets, and amending Regulations (EU) No 1093/2010 and (EU) No 1095/2010 and Directives 2013/36/EU and (EU) 2019/1937 (Text with EEA relevance) (2023) — link
  71. Some stylized facts of the Bitcoin market (2017) — link
  72. Some Traders Are Talking Up Cryptocurrencies, Then Dumping Them, Costing Others Millions — link
  73. Temporal Graph Benchmark for Machine Learning on Temporal Graphs (2023)
  74. The benefits and potential costs of a digital economy (2023) — link
  75. The Economics of Cryptocurrency Pump and Dump Schemes (2019) — link
  76. There’s no free lunch, even using bitcoin: Tracking the popularity and profits of virtual currency scams (2015) — link
  77. Twitter and cryptocurrency pump-and-dumps (2024) — link
  78. When Bitcoin encounters information in an online forum: Using text mining to analyse user opinions and predict value fluctuation (2017) — link
  79. Cloudburst (2024)
  80. Kaiko (2024)
  81. Sui (2024)
  82. STOJR (2024)
  83. Telegram (2024)
  84. Discord (2024)
  85. Twitter (2024)
  86. Reddit (2024)
  87. The Block (2024)
  88. Powerledger (2024)
  89. ERC-20 TOKEN STANDARD (2024)
  90. BEP-20 (2024)
  91. Crypto-Influencer Liability (2024)
  92. 54% of ERC-20 Tokens Listed on DEXes in 2023 Display Patterns That May Be Suggestive of Pump and Dump Schemes, but Represent just 1.3% of DEX Trading Volume (2024)

0 of 100 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.