{"paper_id":"2503.01686","title":"Perseus: Tracing the Masterminds Behind Cryptocurrency Pump-and-Dump Schemes","authors":["Honglin Fu","Yebo Feng","Cong Wu","Jiahua Xu"],"paper_date":"2025-03-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2503.01686v1","page_url":"https://thequant.space/flowcharts/perseus-tracing-the-masterminds-behind-cryptocurre/","scores":{"empirical_rigor":8.5,"math_complexity":2.5,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"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.","quote":"In this research, we develop a detection system PERSEUS, which collects real-time data from the OSN and cryptocurrency markets.","summary":"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…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"At least one node in each temporal attributed graph is a mastermind organizing crowd‑pump events.","type":"model","stated":"explicit","quote":"We assume that at least one node, vm ∈V (Dx τ ), is the mastermind organizing crowd-pump events."},{"assumption":"The maximum return is measured within a three‑day window after the announcement because crowd‑pump effects typically do not last longer.","type":"data","stated":"explicit","quote":"The three‑day window is chosen because crowd‑pump effects typically do not last longer."}]},"datasets":{"status":"available","items":[{"name":"Binance cryptocurrency market data","provider":"Binance","kind":"market-prices","access":"public","period":"February 16 to October 9 2024","universe":"cryptocurrencies","quote":"PERSEUS successfully detects 438 masterminds who are efficient in the pump-and-dump information diffusion networks.","known_id":"binance"},{"name":"Telegram OSN messages","provider":"Telegram","kind":"text-news","access":"public","period":"February 16 to October 9 2024","universe":"crowd‑pump channels","quote":"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."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"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.","kind":"predictive-accuracy","quote":"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)."},{"finding":"GraphSAGE outperforms GAT in precision and F1 score across threshold ranges, with precision peaking around 0.55 and F1 around 0.5.","kind":"predictive-accuracy","quote":"GraphSAGE outperforms GAT in precision within the 0.50–0.6 threshold range, with similar performance elsewhere. The F1 score"},{"finding":"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.","kind":"performance","quote":"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."},{"finding":"The system detected 438 mastermind entities in real‑world deployment.","kind":"other","quote":"PERSEUS successfully detects 438 masterminds who are efficient in the pump-and-dump information diffusion networks."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":46,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:43:45Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}