Scholarlink

<p>Malware continues to evolve through obfuscation and packing, challenging static detectors and limiting the effectiveness of conventional vision-based CNNs that do not jointly capture multi-scale saliency and sequential dependencies encoded in malware images. We propose ParaECA-LSTMNet, a purpose-driven integration of established components: four parallel CNN branches to learn complementary spatial features, an Efficient Channel Attention (ECA) module to emphasize informative channels without added dimensionality, and an LSTM layer to model long-range dependencies before classification. On the Malimg dataset (9,339 grayscale images, 25 families), images are resized to 224 × 224, normalized, and split using a stratified 70%/15%/15% train/validation/test protocol. ParaECA-LSTMNet attains 99.23% Accuracy with Precision/Recall/F1 = 99.23%/99.20%/99.20%, outperforming strong baselines including EfficientNet-B0 (97.43%), VGG-16 + MobileNetV2 (94.57%), Inception-ResNet-V2 (93.23%), and Inception-V3 (92.52%). Confusion-matrix analysis shows minimal inter-family confusion, indicating balanced precision and recall across classes. While the method does not introduce a new primitive, its integration of parallel multi-scale extraction, lightweight channel recalibration, and sequence modeling delivers state-of-the-art accuracy with a compact footprint suitable for deployment as a static pre-filter or in resource-constrained environments.</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

Cybersecurity oriented malware classification using ParaECA-LSTMNet with a hybrid attention guided CNN-LSTM framework

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月8日

查看全文
  • Roise Uddin,
  • Mohammad Mahmudur Rahman,
  • Hossain Ahmed,
  • Amir Hossain Fahad,
  • Aktarun Nesa,
  • Mazharul Islam Arman,
  • Chala Wata

摘要

Malware continues to evolve through obfuscation and packing, challenging static detectors and limiting the effectiveness of conventional vision-based CNNs that do not jointly capture multi-scale saliency and sequential dependencies encoded in malware images. We propose ParaECA-LSTMNet, a purpose-driven integration of established components: four parallel CNN branches to learn complementary spatial features, an Efficient Channel Attention (ECA) module to emphasize informative channels without added dimensionality, and an LSTM layer to model long-range dependencies before classification. On the Malimg dataset (9,339 grayscale images, 25 families), images are resized to 224 × 224, normalized, and split using a stratified 70%/15%/15% train/validation/test protocol. ParaECA-LSTMNet attains 99.23% Accuracy with Precision/Recall/F1 = 99.23%/99.20%/99.20%, outperforming strong baselines including EfficientNet-B0 (97.43%), VGG-16 + MobileNetV2 (94.57%), Inception-ResNet-V2 (93.23%), and Inception-V3 (92.52%). Confusion-matrix analysis shows minimal inter-family confusion, indicating balanced precision and recall across classes. While the method does not introduce a new primitive, its integration of parallel multi-scale extraction, lightweight channel recalibration, and sequence modeling delivers state-of-the-art accuracy with a compact footprint suitable for deployment as a static pre-filter or in resource-constrained environments.