Scholarlink

With the rapid evolving of cybersecurity, the detection of malware and classification of malware have become vital. This work introduces a newly created custom CNN model build aiming to better detection of malware. It’s efficiency is compared with some existing deep learning architectures—VGG19, ResNet50, MobileNetV3—in the context of malware detection. The Malimg dataset, which contains various families of malware images, was used for training and testing this model. We implemented and evaluated these Convolutional Neural Network (CNN) architectures to determine their performance in accurately classifying malware. Our experiment shows that the proposed model, hereby referred to as Modified Convolutional Neural Networks (MCNN), has achieved 99.78% training accuracy and 95.57% testing accuracy within 10 epochs. The results indicate that this MCNN performs better than MobileNetV3 and ResNet50, while achieving results comparable to VGG19 in this Malware detection. This paper provides a comparative analysis of the architectures, highlighting the better performance of the MCNN over MobileNetV3 and ResNet50 in Malware 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

Enhanced Image-Based Blended Malware Detection of 31 Distinct Malware Attacks Using a Modified Deep CNN Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Panchananam Lakshmi Srinivas,
  • Thinnavalli Sree,
  • Surendiran Balasubramanian

摘要

With the rapid evolving of cybersecurity, the detection of malware and classification of malware have become vital. This work introduces a newly created custom CNN model build aiming to better detection of malware. It’s efficiency is compared with some existing deep learning architectures—VGG19, ResNet50, MobileNetV3—in the context of malware detection. The Malimg dataset, which contains various families of malware images, was used for training and testing this model. We implemented and evaluated these Convolutional Neural Network (CNN) architectures to determine their performance in accurately classifying malware. Our experiment shows that the proposed model, hereby referred to as Modified Convolutional Neural Networks (MCNN), has achieved 99.78% training accuracy and 95.57% testing accuracy within 10 epochs. The results indicate that this MCNN performs better than MobileNetV3 and ResNet50, while achieving results comparable to VGG19 in this Malware detection. This paper provides a comparative analysis of the architectures, highlighting the better performance of the MCNN over MobileNetV3 and ResNet50 in Malware detection.