Scholarlink

The increase of the Internet of Things (IoT) ecosystem demands robust security mechanisms to address emerging cyber threats. This paper evaluates the performance of Machine Learning (ML) and Federated Learning (FL) approaches on the BoTNeTIoT-L01 dataset, which represents network traffic from various IoT devices under cyberattack scenarios. Findings demonstrate that while traditional ML models provide high accuracy, reaching up to 99.99%, FL approaches excel in preserving privacy and maintaining robust performance metrics. Even under adversarial conditions such as data poisoning attacks, FL methods show resilience, with accuracy levels sustaining above 90% in non-extreme cases. This study highlights the potential of FL in advancing secure, scalable frameworks for IoT security.

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

Evaluation of Federated Learning for Robust IoT Security Against Label Flipping Attacks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Maryam Alsereidi,
  • Zeyar Aung,
  • Panos Liatsis

摘要

The increase of the Internet of Things (IoT) ecosystem demands robust security mechanisms to address emerging cyber threats. This paper evaluates the performance of Machine Learning (ML) and Federated Learning (FL) approaches on the BoTNeTIoT-L01 dataset, which represents network traffic from various IoT devices under cyberattack scenarios. Findings demonstrate that while traditional ML models provide high accuracy, reaching up to 99.99%, FL approaches excel in preserving privacy and maintaining robust performance metrics. Even under adversarial conditions such as data poisoning attacks, FL methods show resilience, with accuracy levels sustaining above 90% in non-extreme cases. This study highlights the potential of FL in advancing secure, scalable frameworks for IoT security.