Scholarlink

<p>As the Internet develops and the utilization rate of computers increases, the threats posed by malware keep increasing. As the number of newly developed malware is rapidly increasing, the demand for an automatic system for analyzing large volumes of malware also grows together. In this paper, we introduce a technique to automatically analyze malware targeting the Windows environment using a deep learning algorithm. Our proposed method employs convolutional neural network algorithm, which is highly effective in image classification, to detect malware represented as images. To reflect the semantic information of malware for detection, our method uses the opcode frequency data of binary files to generate images while transforming frequency data into spatial data. After calculating the correlation degree among opcodes of malware samples, our method performs clustering using the correlation values as the distances between opcodes. The opcodes in the same cluster are located in a close region in each image, and a CNN model analyzes the generated image. In our evaluation, we conducted the experiments using a large dataset consisting of 10,000 malicious files and 10,000 benign files, and our proposed method was verified to achieve 91% accuracy rate in detecting malicious files.</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

N-gram opcode frequency based malware detection using CNN algorithm

  • 期刊论文

  • 发表日期：

    2025年4月1日

查看全文
  • Seok Min Ko,
  • JaeHyeok Yang,
  • TaeGuen Kim,
  • Ilsun You

摘要

As the Internet develops and the utilization rate of computers increases, the threats posed by malware keep increasing. As the number of newly developed malware is rapidly increasing, the demand for an automatic system for analyzing large volumes of malware also grows together. In this paper, we introduce a technique to automatically analyze malware targeting the Windows environment using a deep learning algorithm. Our proposed method employs convolutional neural network algorithm, which is highly effective in image classification, to detect malware represented as images. To reflect the semantic information of malware for detection, our method uses the opcode frequency data of binary files to generate images while transforming frequency data into spatial data. After calculating the correlation degree among opcodes of malware samples, our method performs clustering using the correlation values as the distances between opcodes. The opcodes in the same cluster are located in a close region in each image, and a CNN model analyzes the generated image. In our evaluation, we conducted the experiments using a large dataset consisting of 10,000 malicious files and 10,000 benign files, and our proposed method was verified to achieve 91% accuracy rate in detecting malicious files.