Using dispersed data and training, federated learning (FL) moves AI capabilities to edge devices or does tasks locally. Many consider FL the start of a new era in AI, yet it is still immature. FL has not garnered the community's trust since its security and privacy implications are controversial. FL's security and privacy concerns must be discovered, analyzed, and recorded before widespread usage and adoption. A solid comprehension of risk variables allows an FL practitioner to construct a secure environment and provide researchers with a clear perspective of potential study fields, making FL the best solution in situations where security and privacy are primary issues. This research aims to deliver a complete overview of FL's security and privacy features to help bridge the gap between current federated AI and broad adoption in the future. In this paper, we present a comprehensive overview of the attack surface to investigate FL's existing challenges and defense measures to evaluate its robustness and reliability. According to our study, security concerns regarding FL are more frequent than privacy issues. Communication bottlenecks, poisoning, and backdoor attacks represent FL's privacy's most significant security threats. In the final part, we detail future research that will assist FL in adapting to real-world settings.

翻译：利用分散的数据和训练方式，联邦学习将人工智能能力迁移至边缘设备或在本地执行任务。许多人认为联邦学习开启了人工智能的新纪元，但该技术仍不成熟。由于其在安全与隐私方面的争议性影响，联邦学习尚未赢得业界的信任。在联邦学习广泛投入使用和普及之前，其安全与隐私问题必须被发现、分析并记录。对风险因素有坚实的理解，能使联邦学习实践者构建安全环境，并为研究人员提供潜在研究领域的清晰视角，从而让联邦学习在安全与隐私成为核心关切的情境中成为最佳解决方案。本研究旨在全面概述联邦学习的安全与隐私特性，以帮助弥合当前联邦人工智能与未来广泛采纳之间的差距。本文全面梳理了攻击面，以探究联邦学习现有的挑战和防御措施，从而评估其鲁棒性和可靠性。根据我们的研究，关于联邦学习的安全问题比隐私问题更为常见。通信瓶颈、投毒攻击和后门攻击是联邦学习隐私面临的最重大安全威胁。在最后部分，我们详细阐述了未来能够帮助联邦学习适应现实场景的研究方向。