Occupational fraud significantly impacts company finances and reputation. Enterprise Resource Planning (ERP) systems store extensive data that can be used for fraud detection, though legitimate transactions dominate the records, making fraud detection a challenging task. Companies should share findings on evolving fraud patterns to stay current and combat fraud collaboratively. However, data security and privacy concerns hinder the sharing of relevant ERP data and information on fraud incidents. Federated Learning (FL) offers a promising solution by enabling the exchange of model parameters instead of sensitive data. This study evaluates the effectiveness of FL for fraud detection in ERP systems compared to traditional machine learning methods. It considers scenarios involving companies not contributing fraud data, varying proportions of fraud cases, different fraud categories, and data volumes, and their influence on the FL process. Federated training, centralized training, and isolated training are compared, highlighting the potential of FL for enhancing occupational fraud detection in ERP systems.

Enterprise Resource Planning (ERP) systems are used by companies to support and automate business processes. Users need to be granted the necessary permissions to be able to perform their work. Following the principle of least privilege, these permissions shall restrict the access to such information and resources only, which are required to complete the tasks involved. However, even using a well-attuned authorization concept, some users may still misuse the ERP system to enrich themselves. Besides a reduction in profit, companies suffer a loss of reputation and trust from their stakeholders. Furthermore, they may be faced with lawsuits from aggrieved customers or suppliers. Such occupational fraud shall therefore be traced and tracked down. Since all business operations are recorded within an ERP system, a variety of different data sources is available. This paper explores the wealth of data sources found in SAP ERP, the most widespread ERP system from SAP, the world’s leading vendor of ERP systems. It examines, in how far such data sources are suited for fraud detection. Previous literature is surveyed. It turns out that the applicability of the data used in previous work is limited or that no suitable data sources are available for developing and evaluating fraud detection techniques. Therefore, this paper proposes three data sets extracted from SAP ERP. The data sets are made available via GitHub and machine learning techniques are applied to evaluate the adequacy for fraud detection. In addition, the feature importance is examined to increase the transparency of fraud detection.

Enterprise Resource Planning (ERP) systems are used to support and to control the business processes of a company or organization. Such systems integrate the data across the entire company into a complete system that is capable of enhancing the key operations in virtually any department. Commonly, running an ERP system helps companies operate more efficiently, however, this also leads to problems. Employees are able to enrich themselves through insider knowledge or by exploiting incomplete or incorrect permission settings. This is referred to as occupational fraud. Since ERP systems keep records of all executed business activities and log system events and permission checks, they provide a variety of different data sources that can be used to detect occupational fraud. This paper reviews existing literature in the area of fraud detection and fraud cases based on ERP data including business data and system data. Using the most widespread ERP system from SAP, the potential and the suitability of the various data sources with respect to fraud detection is examined.