Scholarlink

<p>The rapid evolution of cyber-attacks has demanded the growth of advanced network security mechanisms to protect sensitive information and the critical infrastructure of cloud systems. So, a novel method is required to mitigate unauthorized access and malicious activities within a network. Hence, this paper devises a Transformer Squeeze Excitation Residual Network with FossaServ Optimization Algorithm (TransSE-ResNet_FSOA), which is a novel Deep Learning (DL)-based intrusion detection technique. This proposed method begins by simulating a cloud environment where attackers inject malicious intrusions into servers. For analyzing these threats, the log files are collected from the database. Then, these log files are normalized using the Sigmoid Normalization. Next, the FossaServ Optimization Algorithm (FSOA), which is a combination of the Serval Optimization Algorithm (SOA) and Fossa Optimization Algorithm (FOA), performs an optimal feature selection for enhancing model efficiency. Finally, the TransSE-ResNet_FSOA is utilized in the intrusion detection phase to optimize Transformer Squeeze Excitation Residual Network (TransSE-ResNet) using the FSOA algorithm for attaining superior accuracy in identifying cloud-based attacks. The TransSE-ResNet_FSOA attained a detection rate of 94.877%, an accuracy of 92.757%, and a False Alarm Rate (FAR) of 0.064% using the BoT-IoT dataset at a 90% learning sample.</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

Cloud-Based Intrusion Detection Using Transformer Squeeze Excitation Residual Network with FossaServ Optimization

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Rohith Bhat Chandrasekhar,
  • Praghash Kumaresan,
  • Dangety Sowjanya,
  • Jeevan Nagendra Kumar Yalla

摘要

The rapid evolution of cyber-attacks has demanded the growth of advanced network security mechanisms to protect sensitive information and the critical infrastructure of cloud systems. So, a novel method is required to mitigate unauthorized access and malicious activities within a network. Hence, this paper devises a Transformer Squeeze Excitation Residual Network with FossaServ Optimization Algorithm (TransSE-ResNet_FSOA), which is a novel Deep Learning (DL)-based intrusion detection technique. This proposed method begins by simulating a cloud environment where attackers inject malicious intrusions into servers. For analyzing these threats, the log files are collected from the database. Then, these log files are normalized using the Sigmoid Normalization. Next, the FossaServ Optimization Algorithm (FSOA), which is a combination of the Serval Optimization Algorithm (SOA) and Fossa Optimization Algorithm (FOA), performs an optimal feature selection for enhancing model efficiency. Finally, the TransSE-ResNet_FSOA is utilized in the intrusion detection phase to optimize Transformer Squeeze Excitation Residual Network (TransSE-ResNet) using the FSOA algorithm for attaining superior accuracy in identifying cloud-based attacks. The TransSE-ResNet_FSOA attained a detection rate of 94.877%, an accuracy of 92.757%, and a False Alarm Rate (FAR) of 0.064% using the BoT-IoT dataset at a 90% learning sample.