Gradient boosting models flag TON DEX rug pulls with AUC up to 0.891 using five-minute trading features, but the idle-based definition's high recall is inflated because its one-hour label window overlaps the feature window.
The Kosmosis Approach to Crypto Rug Pull Detection
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
Crypto rug pulls have become a major threat to the integrity of blockchain ecosystems, with illicit activities surging and resulting in significant financial losses. Existing approaches to detect crypto asset fraud are based on the analysis of transaction graphs within blockchain networks. While effective for identifying transaction patterns indicative of fraud, existing approaches do not capture the semantics of transactions and are constrained to blockchain data. Consequently, preventive methods based on transaction graphs are inherently limited. In response to these limitations, we propose the Kosmosis approach, which aims to incrementally construct a knowledge graph as new blockchain and social media data become available. During construction, it aims to extract the semantics of transactions and connects blockchain addresses to their real-world entities by fusing blockchain and social media data in a knowledge graph. This enables novel preventive methods against rug pulls as a form of crypto asset fraud. To demonstrate the effectiveness and practical applicability of the Kosmosis approach, we examine a series of real-world rug pulls. Through this case, we illustrate how Kosmosis can aid in identifying such fraudulent activities by leveraging the insights from the constructed knowledge graph.
citation-role summary
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
REJECT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Detecting Rug Pulls in Decentralized Exchanges: Machine Learning Evidence from the TON Blockchain
Gradient boosting models flag TON DEX rug pulls with AUC up to 0.891 using five-minute trading features, but the idle-based definition's high recall is inflated because its one-hour label window overlaps the feature window.