The role of machines in fraud detection, security, and management control in e-commerce

چکیده

The rapid growth of e-commerce in recent years has brought significant security challenges and financial risks. Traditional fraud detection methods are ineffective against complex and changing patterns of fraudulent activities, making the use of new intelligent technologies essential. This study aimed to provide a comprehensive analysis of the application of machine learning techniques in fraud detection, transaction security, and risk management in e-commerce. The present study, using a systematic review approach and based on the PRISMA protocol, selected and analyzed seven key articles published between 2020 and 2025 from 247 initial articles. The findings show that hybrid machine learning models with 94.3% accuracy in detecting financial statement fraud, an information fusion approach with 32% reduction in operational costs in credit card transactions, and graph neural networks with 87% detection of money laundering patterns in cryptocurrencies have superior performance compared to traditional methods. Also, deep learning and large language models have shown a remarkable ability to dynamically adapt to emerging threats. The main challenges include unbalanced data, model interpretability, and operational implementation, for which deep architectures and data balancing techniques provide effective solutions. The results of this research show that the overall trend is towards hybrid models, deep learning, and attention to operational efficiency, along with technical accuracy. For future research, it is suggested to develop integrated frameworks, improve model interpretability, and conduct case studies in real environments.