LAUSR

As more investors adopt to enter the field of blockchain investment, the Ponzi scheme, a traditional investment scam, has emerged as a hidden fraud in smart contracts. Although some proposed solutions have paid attention to detecting Ponzi schemes in the blockchain, two problems remain: features for detecting Ponzi schemes are incomplete, and algorithms for detecting Ponzi schemes are not sufficiently efficient. Therefore, we innovatively extract the bytecode feature and combine it with user transaction and opcode frequencies to get more comprehensive features. With these features, we propose a smart contract Ponzi scheme identification method based on the improved LightGBM algorithm. Experiments conducted on the real data set of Ethereum prove that our proposed method has improved accuracy dramatically in terms of the $F$ -score index and the AUC index compared with the state-of-the-art methods. In addition, model training speed is improved significantly. Therefore, our method more accurately identifies Ponzi schemes in smart contracts, thus reducing investment risk.