Scholarlink

<p>To overcome the significant challenges posed by new variants of malware to conventional detection techniques, such as low automation, reliance on expert knowledge, and inadequate capability to detect unknown threats in existing memory analysis methods, this paper designs an intelligent detection algorithm for malicious memory segments based on a one-dimensional convolutional network. This algorithm takes raw memory byte sequences as input, referred to as memory segments, and employs a one-dimensional convolutional neural network to automatically learn their deep features and inherent relationships. This approach facilitates an end-to-end automated analysis from data to detection, thereby eliminating the need for complex manual feature engineering. Experimental results show that the designed algorithm achieves a maximum accuracy of 98.28%, precision of 98.94%, recall of 97.6%, F1-score of 0.9826, and AUC value of 0.9972 on the test set, showcasing outstanding detection performance and generalization capability for malicious memory segments. This research offers a novel and effective method for efficient and accurate memory-based malware detection, which can significantly enhance proactive security defense capabilities.</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

Intelligent malware detection method based on memory segments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月4日

查看全文
  • Shilong Yu,
  • Binglong Li,
  • Yong Zhao,
  • Yanru Chen,
  • Yifeng Sun,
  • Heyu Chang

摘要

To overcome the significant challenges posed by new variants of malware to conventional detection techniques, such as low automation, reliance on expert knowledge, and inadequate capability to detect unknown threats in existing memory analysis methods, this paper designs an intelligent detection algorithm for malicious memory segments based on a one-dimensional convolutional network. This algorithm takes raw memory byte sequences as input, referred to as memory segments, and employs a one-dimensional convolutional neural network to automatically learn their deep features and inherent relationships. This approach facilitates an end-to-end automated analysis from data to detection, thereby eliminating the need for complex manual feature engineering. Experimental results show that the designed algorithm achieves a maximum accuracy of 98.28%, precision of 98.94%, recall of 97.6%, F1-score of 0.9826, and AUC value of 0.9972 on the test set, showcasing outstanding detection performance and generalization capability for malicious memory segments. This research offers a novel and effective method for efficient and accurate memory-based malware detection, which can significantly enhance proactive security defense capabilities.