The convergence of edge computing and federated learning (FL) offers promising solutions for privacy-preserving machine learning in Internet of Things (IoT) environments. However, integrating these technologies raises complex challenges in terms of security, communication efficiency, and heterogeneous device capabilities. This survey explores the landscape of secure federated learning techniques tailored for edge-IoT settings, with a focus on differential privacy, secure multiparty computation, homomorphic encryption, and trusted execution environments. It highlights trade-offs between privacy, accuracy, and system performance across various application domains, including smart healthcare, industrial IoT, and autonomous systems. Furthermore, the paper outlines open research challenges such as incentive mechanisms, personalization, model poisoning, and fault tolerance in federated settings. The survey aims to serve as a comprehensive guide for researchers and practitioners striving to build secure, scalable, and efficient FL-based systems in distributed IoT networks.