Scholarlink

<p>In real time scenarios, Virtual machines (VM) in cloud environments are widely used for their augmented resource management that dynamically harmonises computing power, storage, and memory. For identifying ransomware attacks it is necessary to have accurate threat detecting algorithms. For detecting ransomware attacks in cloud environments, a novel ELU-BiLSTM framework is proposed which exploits BiLSTM networks augmented by the ELU activation function. By utilising the discerning function of the Adaptive Jensen-Shannon Divergence Bat Algorithm (AJSD-BA) feature selection reduces irrelevant or redundant attributes. The model is trained with the federated learning (FL) approach allowing the VM for data privacy. The model was implemented in Python and achieves accuracy, precision, recall and F1-Score of 99.2%, 95%, 98%, and 96.5% respectively demonstrating the improved performance with other conventional methods.</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

Adaptive Jensen-Shannon Divergence BA and ELU-BiLSTM: Advanced Techniques for Feature Selection and Cloud Security

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Durai Rajesh Natarajan,
  • Swapna Narla,
  • Sai Sathish Kethu,
  • Sreekar Peddi,
  • Dharma Teja Valivarthi,
  • Purandhar Nandikonda

摘要

In real time scenarios, Virtual machines (VM) in cloud environments are widely used for their augmented resource management that dynamically harmonises computing power, storage, and memory. For identifying ransomware attacks it is necessary to have accurate threat detecting algorithms. For detecting ransomware attacks in cloud environments, a novel ELU-BiLSTM framework is proposed which exploits BiLSTM networks augmented by the ELU activation function. By utilising the discerning function of the Adaptive Jensen-Shannon Divergence Bat Algorithm (AJSD-BA) feature selection reduces irrelevant or redundant attributes. The model is trained with the federated learning (FL) approach allowing the VM for data privacy. The model was implemented in Python and achieves accuracy, precision, recall and F1-Score of 99.2%, 95%, 98%, and 96.5% respectively demonstrating the improved performance with other conventional methods.