Scholarlink

Although Artificial Intelligence (AI) has advanced rapidly in the Internet of Things (IoT), security in distributed model training(whether in industrial applications or other domains) is still a challenge. Federated Learning (FL), a method that enables devices to collaboratively train a shared model while minimizing centralized data exposure, demonstrates superior security properties compared to other methods. However, simply adopting FL cannot guarantee the security of IoT during model training. Therefore, this paper provides a survey on existing approaches to addressing security threats in FL to facilitate its application in IoT. This paper comprehensively categorizes the FL-based approaches dealing with the unique security challenges in different scenarios. Finally, this paper concludes with an analysis of the difficulty when integrating FL with IoT and provide insights for future directions.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Federated Learning for AIoT Security: Current Advances and Future Challenges

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hao Yin,
  • Peng Wang

摘要

Although Artificial Intelligence (AI) has advanced rapidly in the Internet of Things (IoT), security in distributed model training(whether in industrial applications or other domains) is still a challenge. Federated Learning (FL), a method that enables devices to collaboratively train a shared model while minimizing centralized data exposure, demonstrates superior security properties compared to other methods. However, simply adopting FL cannot guarantee the security of IoT during model training. Therefore, this paper provides a survey on existing approaches to addressing security threats in FL to facilitate its application in IoT. This paper comprehensively categorizes the FL-based approaches dealing with the unique security challenges in different scenarios. Finally, this paper concludes with an analysis of the difficulty when integrating FL with IoT and provide insights for future directions.