Scholarlink

<p>Phishing attacks (PAs) remain a major security threat to cloud users, but traditional detection systems are not focused on sharable devices in the cloud environment (CE). To fill this void, this research proposes a sophisticated PA detection framework specifically designed for sharable devices, based on a DBS-BiLSTM model. The framework starts by extracting text contents and URLs from e-mails and datasets and using word embedding to improve text representation. URL-based analysis is performed, such as content and structural feature extraction, with GoogleNet being used for structural feature extraction. In addition, IP grams and SSD features are extracted from the SDD, and IP addresses are one-hot encoded. Feature selection is performed to ensure the most useful attributes are preserved prior to classification with DBS-BiLSTM. The model proposed here greatly improves PA detection accuracy compared to 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

Device-aware phishing attack detection in cloud environments using an enhanced DBS-BiLSTM model

  • 期刊论文

  • 发表日期：

    2025年9月15日

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

摘要

Phishing attacks (PAs) remain a major security threat to cloud users, but traditional detection systems are not focused on sharable devices in the cloud environment (CE). To fill this void, this research proposes a sophisticated PA detection framework specifically designed for sharable devices, based on a DBS-BiLSTM model. The framework starts by extracting text contents and URLs from e-mails and datasets and using word embedding to improve text representation. URL-based analysis is performed, such as content and structural feature extraction, with GoogleNet being used for structural feature extraction. In addition, IP grams and SSD features are extracted from the SDD, and IP addresses are one-hot encoded. Feature selection is performed to ensure the most useful attributes are preserved prior to classification with DBS-BiLSTM. The model proposed here greatly improves PA detection accuracy compared to conventional methods.