Scholarlink

The continuous development and integration of Internet of Things (IoT) devices in today’s generation have also introduced to more significant security vulnerabilities, mainly in rise of cyber-attacks using botnets. Many traditional botnet detection methods have faced various challenges like privacy risks, high latency, high cost of big data transmission due to centralized processing, etc. This research introduced a decentralized IoT botnet detection model that uses Edge-AI, Principal Component Analysis and Federated Learning (FL) in order to address those challenges. In this proposed model, Edge-AI models directly operate at IoT devices that has the network edge, huge dimensionality of the data is then reduced using PCA and enhances the performance of the model in botnet detection real-time while protecting the sensitive data by keeping in local. Implementation of Federated Learning helps in model training across IoT devices by exchanging only model updates also while preserving privacy of the user and minimizing the bandwidth usage. This approach reduces the latency and also enhances the performance in detection and defence against these evolving threats. The proposed model is evaluated on the CICIDS-2018 dataset in order to evaluate the effectiveness of the model in achieving best accuracy in detection with low latency, computational power and privacy preserving solution for botnet detection in IoT landscape.

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

IoT Botnet Detection Through Edge-AI and Federated Learning: A Privacy-Preserving and Low-Latency Approach

  • 章节

  • 发表日期：

    2025年

查看全文
  • Pratik Kumar Swain,
  • Suneeta Satpathy,
  • Deepthi Godavarthi

摘要

The continuous development and integration of Internet of Things (IoT) devices in today’s generation have also introduced to more significant security vulnerabilities, mainly in rise of cyber-attacks using botnets. Many traditional botnet detection methods have faced various challenges like privacy risks, high latency, high cost of big data transmission due to centralized processing, etc. This research introduced a decentralized IoT botnet detection model that uses Edge-AI, Principal Component Analysis and Federated Learning (FL) in order to address those challenges. In this proposed model, Edge-AI models directly operate at IoT devices that has the network edge, huge dimensionality of the data is then reduced using PCA and enhances the performance of the model in botnet detection real-time while protecting the sensitive data by keeping in local. Implementation of Federated Learning helps in model training across IoT devices by exchanging only model updates also while preserving privacy of the user and minimizing the bandwidth usage. This approach reduces the latency and also enhances the performance in detection and defence against these evolving threats. The proposed model is evaluated on the CICIDS-2018 dataset in order to evaluate the effectiveness of the model in achieving best accuracy in detection with low latency, computational power and privacy preserving solution for botnet detection in IoT landscape.