Federated Learning Technology for Data Confidentiality

This technical report discusses federated learning technology and its applicability to generative AI, particularly large language models (LLMs). The document outlines the challenges of data acquisition in AI development, emphasizing the importance of data confidentiality and privacy. It introduces federated learning as a solution that allows organizations to collaborate without centralizing sensitive data, thus mitigating risks of data misuse and leakage. The report details three types of federated learning: horizontal, vertical, and transfer federated learning, each suited for different data scenarios. Horizontal federated learning involves participants with similar data formats, while vertical federated learning accommodates participants with different data types sharing information about the same samples. Transfer federated learning is highlighted for its effectiveness in cross-industry collaborations where data overlap occurs. The report concludes by discussing the usability and challenges of implementing federated learning in generative AI applications.