Scholarlink

Advanced Persistent Threat (APT) attacks are sophisticated cyber threats that target organizations through prolonged, stealthy operations. Despite numerous detection methods, traditional approaches struggle with imbalanced data and high false alarm rates. This paper introduces BiAS, a novel detection model combining Bidirectional Long Short-Term Memory (BiLSTM), Attention mechanism, and Synthetic Minority Over-sampling Technique (SMOTE). BiLSTM analyzes network flows bidirectionally, while Attention highlights significant patterns, and SMOTE addresses data imbalance by generating synthetic attack samples. Experimental results demonstrate BiAS's superior performance over traditional methods, offering a reliable solution for real-time APT detection.

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

BiAS: Transforming Cyber Defense - A Deep Learning Approach for APT Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nguyen Hoa Cương,
  • Do Xuan Cho,
  • Tran Quang Anh

摘要

Advanced Persistent Threat (APT) attacks are sophisticated cyber threats that target organizations through prolonged, stealthy operations. Despite numerous detection methods, traditional approaches struggle with imbalanced data and high false alarm rates. This paper introduces BiAS, a novel detection model combining Bidirectional Long Short-Term Memory (BiLSTM), Attention mechanism, and Synthetic Minority Over-sampling Technique (SMOTE). BiLSTM analyzes network flows bidirectionally, while Attention highlights significant patterns, and SMOTE addresses data imbalance by generating synthetic attack samples. Experimental results demonstrate BiAS's superior performance over traditional methods, offering a reliable solution for real-time APT detection.