Vertical Federated Learning: Principles, Applications, and Future Frontiers

This chapter explores vertical federated learning (VFL), a paradigm that diverges from traditional horizontal FL by vertically partitioning data features and labels across entities. It starts with an introduction, providing context for VFL within the broader landscape of Federated Learning. Highlighting the differences between HFL and VFL architecture. Delving into the principles of VFL, we uncovered the basic pipeline of VFL and various stages of this platform. Some of the main applications of VFL are explored to showcase its unique suitability for scenarios involving dependent entities, such as organizations with shared user data. The challenges and limitations of VFL were thoroughly examined. Delving into studies aimed at mitigating VFL complexities, within these challenges emerges a call for innovation. The chapter explored emerging fields of solutions and innovations in VFL, acknowledging the need for dedicated benchmark tools, and many research opportunities to bridge existing gaps in theory and implementation.