Can We Determine Whether a Set of Ethereum Transaction Data Contains Fraudulent Transactions?

As the demand for cryptographic assets increases, so does the number of fraudulent transactions, necessitating efficient detection methods. In this paper, we propose a method to determine whether a set of transaction data contains fraudulent transactions. We apply topological data analysis, which characterizes the geometric structure of the data, to Ethereum, one of the crypto assets. Our aim is to solve the imbalance in the transaction data used in machine learning models for fraudulent transaction detection. Our method achieved an F1 score of 0.9891 on a set of transaction data containing 10 fraudulent transactions out of 10000 transactions.