Scholarlink

Technologies associated with the Internet of Things (IoT) has continued to advance and evolve with ongoing developments in wide range of academic, economy, healthcare, transportation and communications. With the emerging of advanced technologies, security and privacy is playing a crucial role in driving innovation of IoT. In this paper, it proposes a new dimensional feature reduction, Hybrid Feature Dimension Reduction (HFDR) method for decreasing the number of dimensions of heterogeneous IoT network data based on Denoising Auto-encoder (DAE) and eXtreme Gradient Boosting (XGBoost) and an effective HFDR-BLSTM IDS model through the bench mark IoT intrusion dataset BoT-IoT. Proposed method is also compared with Mutual Information (MI) based feature dimension reduction technique. The proposes HFDR method and IDS model exhibited enhance performance in comparison to the other models. This achievement establishes a highly effective IDS with an accuracy of exceeding 99.95% in the binary classification for detecting IoT network intrusion.

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

Detection of Malicious Activities in IoT Network Based on Enhanced Dimensional Feature Reduction and BLSTM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yee Mon Thant,
  • Zin Thu Thu Myint,
  • Chaw Su Htwe

摘要

Technologies associated with the Internet of Things (IoT) has continued to advance and evolve with ongoing developments in wide range of academic, economy, healthcare, transportation and communications. With the emerging of advanced technologies, security and privacy is playing a crucial role in driving innovation of IoT. In this paper, it proposes a new dimensional feature reduction, Hybrid Feature Dimension Reduction (HFDR) method for decreasing the number of dimensions of heterogeneous IoT network data based on Denoising Auto-encoder (DAE) and eXtreme Gradient Boosting (XGBoost) and an effective HFDR-BLSTM IDS model through the bench mark IoT intrusion dataset BoT-IoT. Proposed method is also compared with Mutual Information (MI) based feature dimension reduction technique. The proposes HFDR method and IDS model exhibited enhance performance in comparison to the other models. This achievement establishes a highly effective IDS with an accuracy of exceeding 99.95% in the binary classification for detecting IoT network intrusion.