Scholarlink

<p>Cloud computing has become essential for organizations to efficiently store, process, and share data while accessing diverse services. However, the increasing volume of data stored and processed in the cloud exposes it to significant security risks, including unauthorized access and cyber threats. Therefore, this study introduces a novel Hybrid Red Panda Simulated Feature Selection with a Machine Learning-Based Intrusion Detection method for enhancing security in cloud infrastructure. The model collects network traffic data from diverse datasets, including UNSW-NB15, Edge IIoT, TON-IoT, NSL-KDD, Cryptojacking attack time series, and BoT-IoT. To address class imbalance, these datasets are balanced using the Synthetic Minority Over-sampling Technique. The steps taken during preprocessing, such as cleaning the data, applying one-hot encoding, and performing Z-score normalization, is crucial for providing high-quality data. The proposed hybrid optimization method combines Red Panda Optimizer and Simulated Annealing to select optimal features, reducing computational complexity and improving detection efficiency. An Ensemble-based Gradient Boosting Regression Tree is employed for anomaly detection, fine-tuned through grid search to achieve robust performance. To enhance decision-making transparency, Shapley Additive Explanations and Local Interpretable Model-Agnostic Explanations are utilized, offering feature-level and instance-specific insights. A comprehensive evaluation of the proposed framework significantly outperforms existing methods, achieving an accuracy of 99.6% and a precision of 99.35%, demonstrating superior reliability. This work provides a robust and interpretable approach to enhancing cloud security and offers a scalable solution for mitigating cyber threats in diverse cloud 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 security in cloud computing systems using hybrid feature selection and ensemble-based machine learning for intrusion detection

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • J. Aswini,
  • K. Sashi Rekha,
  • R. Anto Arockia Rosaline,
  • A. Sivaneshkumar

摘要

Cloud computing has become essential for organizations to efficiently store, process, and share data while accessing diverse services. However, the increasing volume of data stored and processed in the cloud exposes it to significant security risks, including unauthorized access and cyber threats. Therefore, this study introduces a novel Hybrid Red Panda Simulated Feature Selection with a Machine Learning-Based Intrusion Detection method for enhancing security in cloud infrastructure. The model collects network traffic data from diverse datasets, including UNSW-NB15, Edge IIoT, TON-IoT, NSL-KDD, Cryptojacking attack time series, and BoT-IoT. To address class imbalance, these datasets are balanced using the Synthetic Minority Over-sampling Technique. The steps taken during preprocessing, such as cleaning the data, applying one-hot encoding, and performing Z-score normalization, is crucial for providing high-quality data. The proposed hybrid optimization method combines Red Panda Optimizer and Simulated Annealing to select optimal features, reducing computational complexity and improving detection efficiency. An Ensemble-based Gradient Boosting Regression Tree is employed for anomaly detection, fine-tuned through grid search to achieve robust performance. To enhance decision-making transparency, Shapley Additive Explanations and Local Interpretable Model-Agnostic Explanations are utilized, offering feature-level and instance-specific insights. A comprehensive evaluation of the proposed framework significantly outperforms existing methods, achieving an accuracy of 99.6% and a precision of 99.35%, demonstrating superior reliability. This work provides a robust and interpretable approach to enhancing cloud security and offers a scalable solution for mitigating cyber threats in diverse cloud environments.