Detection and Analysis of Cryptocurrency Scams on Twitter

With the increasing adoption of cryptocurrencies in finance, financial fraud has surged, resulting in significant monetary losses. This paper presents a novel approach leveraging advanced graph algorithms to detect fraudulent content in cryptocurrency-related tweets. Our primary objective is to identify potential scam tweets and analyze the patterns within the user communities associated with these tweets. We employ the Label Propagation Algorithm on a unique dataset from Twitter (now X) to detect potential scam tweets and validate our results using the Silhouette score. Following the identification of scam tweets, we analyze the associated user communities using the Louvain Community Detection algorithm. To assess our methodology, we utilize validation metrics such as Modularity, Coverage, and Performance. This comprehensive study aims to identify scam tweets, uncover their unique characteristics, compare them with other fraudulent accounts, and explore connections to broader scam networks. Our proposed framework offers a robust foundation for detecting and understanding fraudulent activities within cryptocurrency discussions on Twitter, achieving a high Silhouette Score for scam tweet identification and strong performance in community detection.