Scholarlink

Background <p>The rapid growth of the Internet of Things (IoT) and pervasive Internet connectivity has led to massive volumes of streaming data in cloud data centers, making real-time anomaly detection a crucial security challenge. Existing anomaly detection techniques often suffer from semantic information loss, high computational complexity, and limited detection accuracy when handling high-dimensional network traffic data.</p> Methods <p>This paper proposes a Semantic-Preserved Generative Adversarial Network with Portia Spider Optimization for Anomaly Detection in Cloud Datacenter Security (SPGAN-PSO-AD-CDS). Network traffic data are obtained from the KDD Cup 99 dataset and preprocessed using Region-Aware Neural Graph Collaborative Filtering (RANGCF) to remove irrelevant information. Feature extraction is performed using the Iterative Matching Synchrosqueezing Transform (IMST) to derive statistical features such as contrast, entropy, standard deviation, and variance. The Seasonal Optimization Algorithm (SOA) is employed for optimal feature selection. The selected features are then classified using a Semantic-Preserved Generative Adversarial Network (SPGAN), while the Portia Spider Optimization Algorithm (PSOA) is applied to fine-tune the network weight parameters.</p> Results <p>The proposed SPGAN-PSO-AD-CDS framework is evaluated using accuracy, F1-score, error rate, detection rate, execution time, and ROC metrics. Experimental results demonstrate that the proposed model achieves 6.86%, 9.75%, and 8.18% higher accuracy compared to ADCC-KGE-ML, AD-TAAI-CC, and software-defined networking– based anomaly detection models, respectively.</p> Conclusion <p>The results confirm that the proposed SPGAN-PSO-AD-CDS framework provides a robust, efficient, and high-accuracy solution for real-time anomaly detection in cloud data center security, outperforming existing state-of-the-art approaches.</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

Semantic- Preserved Generative Adversarial Network with Portia Spider Optimization for Anomaly Detection in Cloud Datacenter Security

  • 期刊论文

  • 发表日期：

    2026年3月20日

查看全文
  • G. Dinesh,
  • D. J. Daniel,
  • S. Aghalya,
  • M. Elangovan

摘要

Background

The rapid growth of the Internet of Things (IoT) and pervasive Internet connectivity has led to massive volumes of streaming data in cloud data centers, making real-time anomaly detection a crucial security challenge. Existing anomaly detection techniques often suffer from semantic information loss, high computational complexity, and limited detection accuracy when handling high-dimensional network traffic data.

Methods

This paper proposes a Semantic-Preserved Generative Adversarial Network with Portia Spider Optimization for Anomaly Detection in Cloud Datacenter Security (SPGAN-PSO-AD-CDS). Network traffic data are obtained from the KDD Cup 99 dataset and preprocessed using Region-Aware Neural Graph Collaborative Filtering (RANGCF) to remove irrelevant information. Feature extraction is performed using the Iterative Matching Synchrosqueezing Transform (IMST) to derive statistical features such as contrast, entropy, standard deviation, and variance. The Seasonal Optimization Algorithm (SOA) is employed for optimal feature selection. The selected features are then classified using a Semantic-Preserved Generative Adversarial Network (SPGAN), while the Portia Spider Optimization Algorithm (PSOA) is applied to fine-tune the network weight parameters.

Results

The proposed SPGAN-PSO-AD-CDS framework is evaluated using accuracy, F1-score, error rate, detection rate, execution time, and ROC metrics. Experimental results demonstrate that the proposed model achieves 6.86%, 9.75%, and 8.18% higher accuracy compared to ADCC-KGE-ML, AD-TAAI-CC, and software-defined networking– based anomaly detection models, respectively.

Conclusion

The results confirm that the proposed SPGAN-PSO-AD-CDS framework provides a robust, efficient, and high-accuracy solution for real-time anomaly detection in cloud data center security, outperforming existing state-of-the-art approaches.