Benchmarking and Opportunities of AI Technologies for Fraud Prevention in the Financial Sector

Abstract

Artificial intelligence (AI) technologies cover key areas of financial sector security, including combating money laundering and fraud, collecting security data, monitoring and preventing cyber threats. The relevance of this study is explained by a lack of research on the implementation and use of AI and comparing individual AI subtechnologies, which makes it difficult to assess their effectiveness, speed and accuracy of recognizing fraudulent schemes. The goal of our research was to assess the possibilities of using various AI technologies to identify suspicious activity and anomalies and to conduct a comparative analysis of their effectiveness in combating fraud in the financial sector. The methods employed included general scientific methods of theoretical knowledge – drawing analogies, induction and deduction, comparative analysis, monographic analysis, and case study. We paid special attention to the experience of the US financial sector. As an empirical base, we used statistical data from Al-Kindi Research and Development Center. The article defines the capabilities of AI technologies: 1) at various stages of fraud prevention according to the CIMA model 2) for various types of financial fraud. According to the results of the study, the efficiency of various models ranges from 88 to 94 %, which indicates their good adaptability to various fraud detection scenarios. Machine learning subtechnologies (decision trees) and deep learning models (neural networks and convolutional neural networks) demonstrate the highest completeness of information disclosure. The positive impact of AI technologies on the fraud detection procedure is an increase in the accuracy of fraud detection by 85 %, an increase in the speed of identifying new fraud schemes by 78 %, an increase in fraud detection by 70 %, and a decrease in false positives by 92 %. We offered the following solutions for the development of AI and the prevention of fraud in the financial sector: data preparation at the collection stage, informed selection of AI models, testing the integration capabilities of AI models, and cooperation with IT companies.