Scholarlink

<p>The Internet of Things (IoT) is growing substantially in today’s digital world. Network security often plays a vital role in IoT networks, yet it has been limited by several cyber threats including generic, worms, fuzzers, Denial of Service (DoS) and so on. Recently, various deep learning (DL) based approaches have been developed to detect and prevent cyber attacks affecting IoT networks but they often failed to detect various types of cyberattacks in IoT environments and showed poor detection accuracy. To overcome these drawbacks, we introduce a novel methodology by integrating a bi-directional long short-term memory-based transformer (BiLSTM-Former), reinforcement learning-enabled graph transformer autoencoder (RL-GTAE) and fuzzy logic-based decision-making techniques for enhancing the security and integrity of IoT networks. The BiLSTM-Former approach incorporates transformer and LSTM architectures for extracting the temporal features. The RL-GTAE model utilized a deep deterministic policy gradient (DDPG) algorithm integrated with a transformer-based graph autoencoder through an adversarial environment for the detection of diverse cyber threats. Moreover, a fuzzy logic-based decision-making technique is deployed for alert generation. The proposed model is executed on the Python 3.9 platform, and the comprehensive performance of the proposed method is measured using UNSW_NB15, CICIDS2017 and NSL-KDD datasets. Our proposed framework attained an accuracy of 98.85%, a recall of 97.92%, a specificity of 97.69%, a precision of 98.56%, a 98.23% F1-score and a 97.74% G-mean score. The results are compared with existing techniques such as SGW-DT, DNN-LSTM, PCA-DNN, GWO-PSO-RF and DCCNN-SMO, and the simulation findings illustrate that our proposed technique obtained outstanding performance in the identification of diverse cyber attacks with higher detection rates. Finally, the experimental results indicate the potentiality of the proposed technique in cyber threat detection and safeguarding IoT networks from a wide range of cyber threats.</p>

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

Engineering solutions for IoT security: a BiLSTM-Former and RL-GTAE-based approach

  • 期刊论文

  • 发表日期：

    2025年7月29日

查看全文
  • GUNASAGARI G S,
  • SHANKARAIAH N,
  • THEJASWINI P

摘要

The Internet of Things (IoT) is growing substantially in today’s digital world. Network security often plays a vital role in IoT networks, yet it has been limited by several cyber threats including generic, worms, fuzzers, Denial of Service (DoS) and so on. Recently, various deep learning (DL) based approaches have been developed to detect and prevent cyber attacks affecting IoT networks but they often failed to detect various types of cyberattacks in IoT environments and showed poor detection accuracy. To overcome these drawbacks, we introduce a novel methodology by integrating a bi-directional long short-term memory-based transformer (BiLSTM-Former), reinforcement learning-enabled graph transformer autoencoder (RL-GTAE) and fuzzy logic-based decision-making techniques for enhancing the security and integrity of IoT networks. The BiLSTM-Former approach incorporates transformer and LSTM architectures for extracting the temporal features. The RL-GTAE model utilized a deep deterministic policy gradient (DDPG) algorithm integrated with a transformer-based graph autoencoder through an adversarial environment for the detection of diverse cyber threats. Moreover, a fuzzy logic-based decision-making technique is deployed for alert generation. The proposed model is executed on the Python 3.9 platform, and the comprehensive performance of the proposed method is measured using UNSW_NB15, CICIDS2017 and NSL-KDD datasets. Our proposed framework attained an accuracy of 98.85%, a recall of 97.92%, a specificity of 97.69%, a precision of 98.56%, a 98.23% F1-score and a 97.74% G-mean score. The results are compared with existing techniques such as SGW-DT, DNN-LSTM, PCA-DNN, GWO-PSO-RF and DCCNN-SMO, and the simulation findings illustrate that our proposed technique obtained outstanding performance in the identification of diverse cyber attacks with higher detection rates. Finally, the experimental results indicate the potentiality of the proposed technique in cyber threat detection and safeguarding IoT networks from a wide range of cyber threats.