Test case prioritization (TCP) is an important part of software testing for maximizing efficiency of the resources and detecting faults as early as possible. The traditional TCP methods use historical data and predefined heuristics which are not flexible enough to adapt to dynamic software environment. In this paper, we suggest a new architecture with federated learning (FL) technologies to improve TCP performance. On one hand, federated learning allows for collaborative training of models across distributed data sources without a need for centralizing sensitive data, which in turn makes it suitable for TCP in CI software environmental setup. In this paper, we propose a comprehensive framework for integrating FL into TCP by considering factors like privacy of data and model aggregation strategy into account. Furthermore, we carry our extensive evaluation on datasets to demonstrate the efficiency of the proposed method with real data. This study proposes that federated learning for TCP can provide a considerable performance boost where 84.2 percentage of the failing test cases is identified and is expected to revolutionize software testing in the future.