Scholarlink

Android malware uses images to display messages and threats, while the malware detection classification uses the AndroidManifest.xml, resources.arsc, and classes.dex files for malware classification and fails to detect such an app. To address this issue, we have presented a DL-based Android malware detection system that uses the features extracted from the three files and the text extracted from the image. In this paper, we have utilized the pre-trained model (Inceptionv3 and MobileNetV2) for feature extraction from RGB Markov images of the app. Then, we collected all the pictures in the app, extracted the text, and performed feature extraction. Our experiment shows that using the text extracted along with the Markov images increases the accuracy. On the Dataset, our accuracy reaches 98.97% without data augmentation compared with the other method.

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

AKdroid: Deep Learning Based-Android Malware Classification Using Markov Images and Text Messages

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kumar Sondarva,
  • Adarsh Kumar,
  • Bhavesh N. Gohil,
  • Devesh C. Jinwala,
  • Sankita J. Patel

摘要

Android malware uses images to display messages and threats, while the malware detection classification uses the AndroidManifest.xml, resources.arsc, and classes.dex files for malware classification and fails to detect such an app. To address this issue, we have presented a DL-based Android malware detection system that uses the features extracted from the three files and the text extracted from the image. In this paper, we have utilized the pre-trained model (Inceptionv3 and MobileNetV2) for feature extraction from RGB Markov images of the app. Then, we collected all the pictures in the app, extracted the text, and performed feature extraction. Our experiment shows that using the text extracted along with the Markov images increases the accuracy. On the Dataset, our accuracy reaches 98.97% without data augmentation compared with the other method.