Scholarlink

<p>Selective Forwarding Attacks (SFAs) threaten Software-Defined Wireless Sensor Networks (SDWSNs) by causing intentional packet drops that degrade reliability, stability, and data integrity. Existing defenses often face issues such as high computational cost, delayed detection under dynamic attacks, lack of post-attack recovery, and absence of privacy-preserving trust management. To overcome these challenges, an Optimized Parallel-head Attention Transformer with Enhanced Northern Goshawk Optimization (OPATNG) is proposed, providing a unified framework for trust encryption, attack detection, and recovery. The process begins with Min–Max and Z-score normalization to stabilize data distribution and suppress outliers. A Prototype CNN (PCNN) then performs deep feature extraction, capturing node behavioral patterns without manual feature design. Extracted features are compressed through a hybrid Error-Controlled Truncated SVD (ECTSVD)–PCA pipeline, reducing dimensionality while retaining discriminative information for efficient edge computation. The OPATNG detection module, optimized via Enhanced Northern Goshawk Optimization (ENGO), enhances convergence, generalization, and anomaly sensitivity across varying attack intensities. For privacy-preserving trust computation, lightweight Fully Homomorphic Encryption (FHE) enables secure aggregation of trust values in the encrypted domain. Finally, the Collaborative Edge–Cloud Trust-Aware Recovery (CETAR) mechanism rapidly isolates compromised nodes and restores stable communication with minimal latency. Experiments on WSN-DS and IRAD datasets show superior performance, achieving 98.1% and 98.7% detection accuracy, 0.5&#xa0;s computation time, 12.3&#xa0;ms latency, and 92% feature retention, outperforming state-of-the-art methods. The proposed OPATNG framework establishes a secure, energy-efficient, and scalable defense architecture for real-time SFA mitigation in next-generation SDWSN and 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

Deep learning-based detection and recovery mechanism for mitigating selective forwarding attacks in event-driven wireless sensor network

  • 期刊论文

  • 发表日期：

    2026年3月4日

查看全文
  • Atul Kumar Agnihotri,
  • Vishal Awasthi

摘要

Selective Forwarding Attacks (SFAs) threaten Software-Defined Wireless Sensor Networks (SDWSNs) by causing intentional packet drops that degrade reliability, stability, and data integrity. Existing defenses often face issues such as high computational cost, delayed detection under dynamic attacks, lack of post-attack recovery, and absence of privacy-preserving trust management. To overcome these challenges, an Optimized Parallel-head Attention Transformer with Enhanced Northern Goshawk Optimization (OPATNG) is proposed, providing a unified framework for trust encryption, attack detection, and recovery. The process begins with Min–Max and Z-score normalization to stabilize data distribution and suppress outliers. A Prototype CNN (PCNN) then performs deep feature extraction, capturing node behavioral patterns without manual feature design. Extracted features are compressed through a hybrid Error-Controlled Truncated SVD (ECTSVD)–PCA pipeline, reducing dimensionality while retaining discriminative information for efficient edge computation. The OPATNG detection module, optimized via Enhanced Northern Goshawk Optimization (ENGO), enhances convergence, generalization, and anomaly sensitivity across varying attack intensities. For privacy-preserving trust computation, lightweight Fully Homomorphic Encryption (FHE) enables secure aggregation of trust values in the encrypted domain. Finally, the Collaborative Edge–Cloud Trust-Aware Recovery (CETAR) mechanism rapidly isolates compromised nodes and restores stable communication with minimal latency. Experiments on WSN-DS and IRAD datasets show superior performance, achieving 98.1% and 98.7% detection accuracy, 0.5 s computation time, 12.3 ms latency, and 92% feature retention, outperforming state-of-the-art methods. The proposed OPATNG framework establishes a secure, energy-efficient, and scalable defense architecture for real-time SFA mitigation in next-generation SDWSN and IoT environments.