The AI system discovered over 95K scam lists between June 2022 and June 2023 that were created from 87,617 accounts on the X (former Twitter) social network
Researchers from San Diego State University in California developed and tested an artificial intelligence (AI) system to search for, track and unmask free cryptocurrency giveaway scams posted on X (ex-Twitter).
The scientists used the GiveawayScamHunter tool to autonomously extract website and wallet addresses associated with fraudulent activities. As a result, the team discovered 95,111 scam lists between June 2022 and June 2023. They were created from 87,617 different accounts on the X social network. Besides, the researchers unearthed 327 scam giveaway internet domains and 121 new scam-related cryptocurrency wallet addresses.
Twitter List is a curated selection of Twitter accounts. Individual users can create up to 1,000 Lists, each with up to 5,000 users on it. The main idea of the feature is to enable business people to organize their marketing strategies, as well as seamlessly track influencers, competitors, customers and more. However, it appears that due to the permissionless nature of Twitter Lists, it has become a simple and popular networking tool for crypto scammers to exploit.
The capabilities of large language models (LLMs) enabled the scientists to determine which lists dealt with giveaway scams. For this purpose, the team trained a natural language processing tool on data from previously identified giveaway scams. As a result, the team managed to compile data on previously unreported scam websites and wallets.
Moreover, the researchers received numerous insights into the scenarios of these scams, targeting methods and the approximate number of victims scammed during the one-year study period.
The experimental study found out that over 365 victims have been attacked by the scammers, with estimated financial losses of $872K.
The team reported the study results and details about the associated accounts, domains and wallet addresses to both X and the cryptocurrency/blockchain community. Nevertheless, according to the paper, 43.9% of the associated accounts still remained active as of Aug. 10. At the same time, the researchers noted that the majority of these spam accounts are likely not in active use.