Scholarlink

<p>Network Control Systems (NCS), including Wireless Sensor Networks (WSN) are broadly deployed across different application areas. A prominent dispute in WSNs is the liability to wormhole attacks, which lead to routing errors, degradation of sensor network lifetime, and disruption of network topology. Despite the development of numerous Wormhole attack detection methods, many of these techniques require additional hardware or consume significant system resources, limiting their practicality. This paper introduces a novel detection framework leveraging the ResNeXt architecture in conjunction with a Deep Stacked Autoencoder (ResNeXt-DSAE) for effective wormhole attack detection in NCSs. The framework begins with the simulation of a WSN, where routing is functioned using the Low Energy Adaptive Clustering Hierarchy (LEACH) protocol. The detection process is structured into multiple phases, starting with the evaluation of the Neighbor Ratio Threshold (NRT), followed by the wormhole attack detection through out-of-band and in-band detection strategies. In the out-of-band detection phase, the transmission range is analyzed, while in-band detection assesses Round-Trip Time (RTT) and Packet Delivery Ratio (PDR). The wormhole attack classification is subsequently performed using the ResNeXt-DSAE framework to distinguish between out-of-band and in-band attacks. Investigational outcomes demonstrate that the devised ResNeXt-DSAE framework accomplishes superior proficiency, with a maximum throughput of 97.55&#xa0;Mbps, Packet Deliver Ratio (PDR) of 99.58%, network lifetime of 0.945, and a minimal delay of 0.815&#xa0;s, network activity energy consumption of 0.255&#xa0;J, and computational cost of 17.99&#xa0;s for 200 nodes. Furthermore, the proposed method attains a detection rate of 0.958, an accuracy of 0.958, a precision of 0.940, a recall of 0.970, an F1-score of 0.955, a False Positive Rate (FPR) of 0.058, and a False Negative Rate (FNR) of 0.077 thereby transcending existing wormhole attack detection 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

ResNeXt and Deep Stacked Autoencoder based framework for wormhole attack detection on network control system

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月31日

查看全文
  • Avina A,
  • Geetha M. R

摘要

Network Control Systems (NCS), including Wireless Sensor Networks (WSN) are broadly deployed across different application areas. A prominent dispute in WSNs is the liability to wormhole attacks, which lead to routing errors, degradation of sensor network lifetime, and disruption of network topology. Despite the development of numerous Wormhole attack detection methods, many of these techniques require additional hardware or consume significant system resources, limiting their practicality. This paper introduces a novel detection framework leveraging the ResNeXt architecture in conjunction with a Deep Stacked Autoencoder (ResNeXt-DSAE) for effective wormhole attack detection in NCSs. The framework begins with the simulation of a WSN, where routing is functioned using the Low Energy Adaptive Clustering Hierarchy (LEACH) protocol. The detection process is structured into multiple phases, starting with the evaluation of the Neighbor Ratio Threshold (NRT), followed by the wormhole attack detection through out-of-band and in-band detection strategies. In the out-of-band detection phase, the transmission range is analyzed, while in-band detection assesses Round-Trip Time (RTT) and Packet Delivery Ratio (PDR). The wormhole attack classification is subsequently performed using the ResNeXt-DSAE framework to distinguish between out-of-band and in-band attacks. Investigational outcomes demonstrate that the devised ResNeXt-DSAE framework accomplishes superior proficiency, with a maximum throughput of 97.55 Mbps, Packet Deliver Ratio (PDR) of 99.58%, network lifetime of 0.945, and a minimal delay of 0.815 s, network activity energy consumption of 0.255 J, and computational cost of 17.99 s for 200 nodes. Furthermore, the proposed method attains a detection rate of 0.958, an accuracy of 0.958, a precision of 0.940, a recall of 0.970, an F1-score of 0.955, a False Positive Rate (FPR) of 0.058, and a False Negative Rate (FNR) of 0.077 thereby transcending existing wormhole attack detection approaches.