Scholarlink

<p>Wireless sensor networks (WSN) have acquired great importance from industries owing to their wide range of applications. In system security, intrusion detection systems (IDS) play a significant role. In WSN, data aggregation protocols are utilized to maximize the network’s life span and minimize every sensor node’s communication weight as well as energy consumption. However, WSN faces energy and resource conservation conflicts. Furthermore, assuming uniform initial energy levels among sensors in existing algorithms is impractical for real-world applications. This variation in initial energy levels has a significant effect on data aggregation within sensor networks. To mitigate these issues, an innovative approach named ShuffleNet Fuzzy Zeiler and Fergus network (Shuffle-F-ZFNet) for data aggregation in WSN is proposed. Primarily, the system model for dynamic cluster WSN is simulated, by considering energy, mobility, link lifetime (LLT) and trust model. Thereafter, energy prediction is conducted utilizing the Deep Neuro-Fuzzy Network (DNFN). Then, dynamic cluster (DC) computation is carried out utilizing the Adaptive Genetic Fuzzy System (AGFS) by considering objectives like distance, throughput, trust factors, LLT, delay, residual and predicted energy. Here, AGFS is trained using Fractional Flamingo Jellyfish Search Optimization (FFJSO), which combines Fractional Calculus (FC) with Flamingo Jellyfish Search Optimization (FJSO). FJSO integrates Flamingo Search Algorithm (FSA) and Jellyfish Search Optimization (JSO). Next, the routing process is executed utilizing FFJSO with the above-mentioned objectives as fitness parameters. At Base Station (BS), feature selection is performed utilizing Angular Separation Distance (ASD) by getting input log data. Then, intrusion attack detection is carried out utilizing a Convolutional Deep Q Learning Network (CDQ-LN), which is designed by integrating a Deep Q-learning Network (DQN) and Convolutional Neural Network (CNN), where layers are modified. Finally, data aggregation based on malicious activity is performed utilizing Shuffle-F-ZFNet. However, Shuffle-F-ZFNet is the combination of ShuffleNet and Zeiler and Fergus network (ZFNet), where layers are modified employing a fuzzy concept with adaptive weightage. Moreover, the proposed Shuffle-F-ZFNet attained the highest energy of 0.958&#xa0;J, distance of 0.103&#xa0;m, delay of 0.084&#xa0;s, trust of 0.193, throughput of 81.355 Mbps and data packet delivery rate of 0.907. Additionally, the proposed method improves performance by 10.12% over Enhanced Energy Optimization Routing Protocol (EEORP), 8.66% over Hybridization of Metaheuristic Cluster-based Routing (HMBCR), 5.63% over Lightweight and Efficient Dynamic Cluster Head Election routing protocol (LEDCHE-WSN), 4.07% over Hybrid and Dynamic Clustering and Routing algorithm (HDCR), 4.07% over FJSO, and 1.56% over FFJSO.</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

Shuffle-F-ZFNet: ShuffleNet Fuzzy Zeiler and Fergus network for data aggregation in WSN data communication

  • 期刊论文

  • 发表日期：

    2025年6月13日

查看全文
  • S. Dhanabal,
  • Chandrashekhar Goswami,
  • Rani Venkata Satya Praveen,
  • T. Vetriselvi

摘要

Wireless sensor networks (WSN) have acquired great importance from industries owing to their wide range of applications. In system security, intrusion detection systems (IDS) play a significant role. In WSN, data aggregation protocols are utilized to maximize the network’s life span and minimize every sensor node’s communication weight as well as energy consumption. However, WSN faces energy and resource conservation conflicts. Furthermore, assuming uniform initial energy levels among sensors in existing algorithms is impractical for real-world applications. This variation in initial energy levels has a significant effect on data aggregation within sensor networks. To mitigate these issues, an innovative approach named ShuffleNet Fuzzy Zeiler and Fergus network (Shuffle-F-ZFNet) for data aggregation in WSN is proposed. Primarily, the system model for dynamic cluster WSN is simulated, by considering energy, mobility, link lifetime (LLT) and trust model. Thereafter, energy prediction is conducted utilizing the Deep Neuro-Fuzzy Network (DNFN). Then, dynamic cluster (DC) computation is carried out utilizing the Adaptive Genetic Fuzzy System (AGFS) by considering objectives like distance, throughput, trust factors, LLT, delay, residual and predicted energy. Here, AGFS is trained using Fractional Flamingo Jellyfish Search Optimization (FFJSO), which combines Fractional Calculus (FC) with Flamingo Jellyfish Search Optimization (FJSO). FJSO integrates Flamingo Search Algorithm (FSA) and Jellyfish Search Optimization (JSO). Next, the routing process is executed utilizing FFJSO with the above-mentioned objectives as fitness parameters. At Base Station (BS), feature selection is performed utilizing Angular Separation Distance (ASD) by getting input log data. Then, intrusion attack detection is carried out utilizing a Convolutional Deep Q Learning Network (CDQ-LN), which is designed by integrating a Deep Q-learning Network (DQN) and Convolutional Neural Network (CNN), where layers are modified. Finally, data aggregation based on malicious activity is performed utilizing Shuffle-F-ZFNet. However, Shuffle-F-ZFNet is the combination of ShuffleNet and Zeiler and Fergus network (ZFNet), where layers are modified employing a fuzzy concept with adaptive weightage. Moreover, the proposed Shuffle-F-ZFNet attained the highest energy of 0.958 J, distance of 0.103 m, delay of 0.084 s, trust of 0.193, throughput of 81.355 Mbps and data packet delivery rate of 0.907. Additionally, the proposed method improves performance by 10.12% over Enhanced Energy Optimization Routing Protocol (EEORP), 8.66% over Hybridization of Metaheuristic Cluster-based Routing (HMBCR), 5.63% over Lightweight and Efficient Dynamic Cluster Head Election routing protocol (LEDCHE-WSN), 4.07% over Hybrid and Dynamic Clustering and Routing algorithm (HDCR), 4.07% over FJSO, and 1.56% over FFJSO.