Federated Learning (FL) has achieved extensive adoption, especially in applications like healthcare and cyber-physical systems, serving as a protective measure for data while ensuring participant privacy. In FL, adversarial attacks present a considerable risk to both the integrity of the learning model and the privacy of the distributed data. The decentralized structure of FL exacerbates this vulnerability, as the data stays local and is not accessible to the central server, complicating efforts to protect against adversarial attacks. This challenge underscores the necessity for further research on robust defense approaches to guarantee that FL can effectively safeguard data privacy and become a viable solution in real-world applications. This article provides an extensive review, including potential attacks and mitigation strategies. This survey presents a taxonomy of adversarial attacks and defense mechanisms, offering a comprehensive overview of the vulnerabilities in FL and the strategies available to mitigate them. Besides, we introduce a unified adversary-resilient FL framework that integrates Blockchain to enhance security. Finally, we present open research challenges in the field of FL.