The rapid increase in data generated by connected devices has created a pressing need for privacy-preserving techniques in distributed learning. This article examines methods that enable collaborative machine learning (ML) while maintaining data security and user privacy. Key approaches such as federated learning (FL), differential privacy (DP), secure multiparty computation (SMPC), and homomorphic encryption (HE) are analyzed for their unique capabilities and various applications. FL facilitates model training across decentralized data sources, ensuring data remain local, while DP mitigates privacy risks by adding controlled noise. SMPC and HE support secure computations on encrypted data, maintaining confidentiality during processing. Despite their effectiveness, these techniques face challenges related to computational complexity, scalability, and regulatory compliance. The article reviews current advancements, practical implementations, and future directions, emphasizing the need for optimized, accessible solutions to enhance data security in distributed systems.