Ethereum Ponzi Scheme Detection Based on PD-SECR

Abstract

Ethereum, a typical application of blockchain technology, has attracted extensive attention from all walks of life since its release. Owing to imperfections in existing supervision technology, illegal and criminal activities on blockchain platforms are becoming increasingly frequent. The most typical Ethereum fraud is the Ponzi scheme, which causes blockchain investors to lose millions of assets and severely impacts social development. Currently, Ponzi scheme detection primarily focuses on machine learning and data mining. However, existing detection methods still have two problems in data imbalance processing and feature extraction: (1) data enhancement using an oversampling algorithm produces noise and (2) feature redundancy existing in extracted feature data. The SMOTEENN algorithm is introduced to solve data imbalance. The PD-SECR method, the Convolutional Neural Network (CNN) feature extraction, and random forest (RF) classification models are used for detection, but the two models are independently trained. The results show that the detection method proposed in this study is more suitable for the Ethereum Ponzi scheme.