Scholarlink

<p>The rapid proliferation of Internet of Things (IoT) devices has introduced significant cybersecurity challenges, particularly due to their limited computational resources and vulnerability to privacy breaches. These devices often operate in sensitive environments, making them prime targets for cyberattacks that aim to compromise personal or organizational data. A key obstacle to improving IoT security lies in the lack of balanced, privacy-compliant, and publicly available datasets necessary to train effective machine learning models. This study aims to address these challenges by exploring the effectiveness of advanced deep learning models, particularly Recurrent Neural Network (RNN)-based Generative Adversarial Networks (GANs), to generate realistic synthetic data to improve IoT security. Specifically, we propose and evaluate multiple GAN architectures, including LSTM-GAN, GRU-GAN, and BiLSTM-GAN, to synthesize high-quality data derived from the real-world UNSW-NB15 dataset. These synthetic data sets are used to train and evaluate various classification models, such as Random Forest (RF), XGBoost (XGB), LightGBM (LGBM) and Decision Tree (DT). Experimental results demonstrate that the GRU-GAN model achieves superior performance in terms of execution time and data realism, while the RF, XGB, LGBM, and DT classifiers achieve 100% precision with reduced training time, highlighting the robustness and efficiency of the proposed framework. In conclusion, this work offers a practical and privacy-conscious solution to the data scarcity problem in IoT cybersecurity by leveraging GAN-generated data, and provides a benchmark for deploying lightweight yet accurate classification models in real-world IoT environments.</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

Enhancing IoT security with advanced GAN architectures for cyberattacks detection

  • 期刊论文

  • 发表日期：

    2025年10月16日

查看全文
  • Mohammed Tayebi,
  • Said El Kafhali

摘要

The rapid proliferation of Internet of Things (IoT) devices has introduced significant cybersecurity challenges, particularly due to their limited computational resources and vulnerability to privacy breaches. These devices often operate in sensitive environments, making them prime targets for cyberattacks that aim to compromise personal or organizational data. A key obstacle to improving IoT security lies in the lack of balanced, privacy-compliant, and publicly available datasets necessary to train effective machine learning models. This study aims to address these challenges by exploring the effectiveness of advanced deep learning models, particularly Recurrent Neural Network (RNN)-based Generative Adversarial Networks (GANs), to generate realistic synthetic data to improve IoT security. Specifically, we propose and evaluate multiple GAN architectures, including LSTM-GAN, GRU-GAN, and BiLSTM-GAN, to synthesize high-quality data derived from the real-world UNSW-NB15 dataset. These synthetic data sets are used to train and evaluate various classification models, such as Random Forest (RF), XGBoost (XGB), LightGBM (LGBM) and Decision Tree (DT). Experimental results demonstrate that the GRU-GAN model achieves superior performance in terms of execution time and data realism, while the RF, XGB, LGBM, and DT classifiers achieve 100% precision with reduced training time, highlighting the robustness and efficiency of the proposed framework. In conclusion, this work offers a practical and privacy-conscious solution to the data scarcity problem in IoT cybersecurity by leveraging GAN-generated data, and provides a benchmark for deploying lightweight yet accurate classification models in real-world IoT environments.