Title: Federated Learning-Based Secure Data Collaboration Across SAP Modules in Cloud Environments

Abstract: While enterprise systems shift towards cloud-native architectures, secure data collaboration is still a challenge within modular components of platforms like SAP S/4HANA. Centralized machine learning models need a data pool from multiple SAP modules, including Finance (FI), Materials Management (MM), Sales and Distribution (SD), which poses privacy, compliance, and system latency risks. This paper presents a solution for secure decentralized intelligence sharing across modules in cloud environments with a federated learning-based framework. The solution enables high model accuracy and robustness while ensuring data sovereignty by allowing localized model training within each module and only aggregating encrypted learning updates. SAP Cloud Connector and Business Technology Platform (BTP) APIs have been augmented to allow seamless integration with the layered privacy design based on secure aggregation and differential privacy. The proposed framework is evaluated across synthetic and real SAP workflow datasets and is shown to achieve up to 92% accuracy while reducing data transfer by 68% and remaining resilient to node failure scenarios. These findings confirm that federated learning is a plausible solution for scalable and privacy-preserving intelligent collaboration in enterprise software ecosystems.

Keywords: Federated Learning in SAP, Secure Cloud Data Collaboration, Privacy-Preserving Enterprise AI.