In recent years, Federated Learning (FL) has emerged as a transformative paradigm for training machine learning models in a decentralized manner, enabling data privacy by keeping data localized on edge devices. However, the distributed nature of FL introduces unique challenges related to privacy and security, which can compromise its effectiveness and adoption in sensitive applications. This survey provides a comprehensive review of existing approaches aimed at enhancing the privacy and security of FL systems. Key topics include differential privacy, secure multi-party computation, homomorphic encryption, and trusted execution environments. Additionally, we explore adversarial threats such as inference, poisoning, model inversion, and backdoor attack, along with countermeasures designed to mitigate these risks. The survey also examines emerging trends in privacy-preserving FL, including hybrid techniques, federated reinforcement learning, blockchain-based solutions, and generative adversarial networks. By synthesizing current research, this article highlights open challenges and future directions for advancing the privacy and security of FL systems.