The rapid expansion of precision agriculture technologies has generated vast quantities of heterogeneous farm-level data, yet traditional centralized data-sharing models face significant barriers including privacy concerns, data sovereignty issues, competitive sensitivities, and regulatory compliance challenges that limit collaborative intelligence development across multiple agricultural stakeholders. Federated learning emerges as a transformative paradigm that enables secure multi-farm collaboration by training shared machine learning models on distributed datasets without requiring raw data exchange, thereby preserving individual farm privacy while enabling collective intelligence. This review examines federated learning frameworks specifically designed for secure agricultural data sharing, analyzing core architectures including horizontal, vertical, and federated transfer learning approaches adapted for farm environments. We explore essential privacy-preserving mechanisms such as secure aggregation protocols, differential privacy techniques, homomorphic encryption, and blockchain-based trust management that protect sensitive farm information during collaborative model training. Key agricultural applications are discussed, including federated crop yield prediction, distributed disease surveillance systems, cross-farm pest forecasting, and collaborative decision-support platforms that leverage multi-stakeholder data while maintaining data ownership. Despite promising advances, significant challenges remain regarding communication efficiency in rural networks, heterogeneity across diverse farming systems, model convergence with non-identically distributed data, and establishment of governance frameworks for federated agricultural intelligence. Successful deployment of federated learning in multi-farm contexts requires interdisciplinary integration of distributed machine learning, agricultural domain knowledge, cybersecurity protocols, and stakeholder engagement strategies to realize scalable privacy-preserving collaborative intelligence for sustainable digital agriculture transformation.