Secure and Adaptive Federated Learning Pipelines: A Framework for Multi-Tenant Enterprise Data Systems
Pooja Devaraju, Shivareddy Devarapalli, Raghavender Reddy Tuniki, Srikanth Kamatala · 2025
Due to the rising amount of data in multi-tenant enterprises, machine learning solutions need to address issues of privacy, scalability, and adaptability. The goal of this paper is to introduce an innovative framework called SAFLP for handling security and adaptability issues in multi-tenant data environments. Having tenant-aware orchestration, dynamic client selection, and adaptive model aggregation makes the framework suitable for any enterprise environment in terms of security and scalability. We implement differential privacy and secure aggregation to reduce the risk of sensitive data being exposed while the model’s accuracy is preserved. As a result, the framework is able to adjust to different resource conditions and the presence of clients, ensuring reliable and effective learning. Tests on actual enterprise data show that the framework outperforms other federated learning methods in terms of accuracy, time required to converge, and ability to meet privacy standards. This study helps pave the way for using federated AI in large enterprises with different data sources.