Scholarlink

<p>Software-defined networking makes modern communication more flexible but also more vulnerable to security concerns. This problem requires detection algorithms that can understand high-dimensional traffic, respond to new behavioral patterns, and mitigate quickly. This book presents an ensemble-driven threat detection framework that employs gradient-boosted decision approaches and recurrent temporal modeling to detect malicious traffic flows in dynamic network circumstances with increased sensitivity. The pipeline standardizes traffic representations for incremental learning via structured feature refinement of raw flow records. The method involves dual-phase analysis. The first phase compares temporal irregularities and protocol violations to identify questionable flows using coarse-grained score filters. Phase two uses a fine-grained classifier to distinguish coordinated attacks from benign anomalies and transient bursts. Each component minimizes low-frequency attack class misclassification and maintains performance under changeable demand. Rapid policy changes from a controller-level enforcement layer enable traffic rerouting, selective rate limitation, and segment isolation. An experimental study examines performance utilizing many intrusion datasets with different attacks. By refining features, inference overhead is lowered while detection accuracy is maintained. Autonomous enforcement of high-priority warnings enhances sensitivity to moderate changes in traffic behavior and reduces delays. Minority attack category recall is better than standard single-model classifiers, and false-positive rates are stable across encrypted and unencrypted flow volumes. Ensemble-based learning, temporal traffic interpretation, and programmable mitigation policies are integrated into an SDN-orchestrated architecture. The solution emphasizes dynamic behavioral modeling and control-plane logic integration over physical-layer optimization and static payload inspection. Cloud-assisted domains, enterprise infrastructures, and distributed edge deployments benefit from scalability. The findings demonstrate a plausible path to responsive, autonomous defense in programmable network systems with consistent accuracy under shifting threat scenarios.</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

Design of an integrated unified intelligence-driven method for low-latency, energy-efficient, and secure cloudintegrated optical broadband access networks

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Antimbala Marmat,
  • Dolly Thankachan

摘要

Software-defined networking makes modern communication more flexible but also more vulnerable to security concerns. This problem requires detection algorithms that can understand high-dimensional traffic, respond to new behavioral patterns, and mitigate quickly. This book presents an ensemble-driven threat detection framework that employs gradient-boosted decision approaches and recurrent temporal modeling to detect malicious traffic flows in dynamic network circumstances with increased sensitivity. The pipeline standardizes traffic representations for incremental learning via structured feature refinement of raw flow records. The method involves dual-phase analysis. The first phase compares temporal irregularities and protocol violations to identify questionable flows using coarse-grained score filters. Phase two uses a fine-grained classifier to distinguish coordinated attacks from benign anomalies and transient bursts. Each component minimizes low-frequency attack class misclassification and maintains performance under changeable demand. Rapid policy changes from a controller-level enforcement layer enable traffic rerouting, selective rate limitation, and segment isolation. An experimental study examines performance utilizing many intrusion datasets with different attacks. By refining features, inference overhead is lowered while detection accuracy is maintained. Autonomous enforcement of high-priority warnings enhances sensitivity to moderate changes in traffic behavior and reduces delays. Minority attack category recall is better than standard single-model classifiers, and false-positive rates are stable across encrypted and unencrypted flow volumes. Ensemble-based learning, temporal traffic interpretation, and programmable mitigation policies are integrated into an SDN-orchestrated architecture. The solution emphasizes dynamic behavioral modeling and control-plane logic integration over physical-layer optimization and static payload inspection. Cloud-assisted domains, enterprise infrastructures, and distributed edge deployments benefit from scalability. The findings demonstrate a plausible path to responsive, autonomous defense in programmable network systems with consistent accuracy under shifting threat scenarios.