DS-Ponzi: Anti-jamming Detection of Ponzi Scheme on Ethereum Utilizing Dynamic-Static Features of Smart Contract Codes

Kun Qian, Jinping Jia, Zhao Zhang, Xiang Li, Yanqin Yang, Cheqing Jin

Abstract: Ponzi scheme contracts on Ethereum have led to substantial economic losses, emphasizing the critical need for their identification. Existing detection methods rely on static features such as bytecode, opcode, and control flow graph (CFG). However, these methods exhibit two shortcomings: (1) Most existing methods only utilize a single static feature, either static opcode or CFG, lacking effective feature fusion and leaving room for improvement in accuracy and recall. (2) Static feature-based methods lack anti-jamming capabilities, as Ponzi scheme designers can inject invalid code to confuse the model’s identification results. To address these issues, we propose a novel Ponzi contract detection model, DS-Ponzi. DS-Ponzi designs a classifier integrating both CFG and opcode features for the classification of Ponzi schemes. Furthermore, DS-Ponzi utilizes dynamic EVM simulation execution to trace the execution paths of functions, thereby pruning CFG and opcode to obtain their dynamic representation. This process effectively captures the core information of Ponzi schemes, enhancing the system’s resilience against code injection attacks. Experimental results demonstrate that DS-Ponzi outperforms existing single-feature methods in both recall and F1-score while enhancing the anti-jamming capability of the detection model.

External IDs:dblp:conf/dasfaa/QianJZLYJ24