Mixed Integer Program for e-Wallet Security Based on Suspicious Transaction Detection

Abstract

With the rise of digital transactions, e-Wallets have become prime targets for fraudulent activity. Early detection of suspicious transactions is therefore crucial to protect users and maintain trust in these systems. This article proposes a mathematical model based on a Mixed Integer Program (MIP) to identify and block suspicious transactions. This mathematical approach is designed to analyze e-Wallet transactions in real time by combining integer and continuous decision variables, offering greater flexibility in modeling fraud detection constraints. It considers parameters such as transaction cost, geographical location, type of device used for the transaction, IP address and other potential fraud indicators. Assigning suspicion scores to each transaction enables the model to identify risks that become habitual behavior and mark them as suspicious. Tests carried out on 10,000 digital transactions show that using PMNE significantly improves the detection of fraudulent transactions by identifying the most critical anomalies in terms of accuracy, adaptability and operational efficiency. The model also offers greater accuracy, reducing the number of false positives and false negatives, enabling faster intervention to block truly suspicious transactions.