Scholarlink

The fast evolution of new malware variants poses a formidable challenge to cybersecurity, underscoring the critical necessity for sophisticated and robust classification techniques. Traditional techniques, which often depend on analyzing disassembled code, face limitations such as susceptibility to anti-disassembly strategies and increased error rates, reducing their efficacy in malware detection. To overcome these challenges, the proposed work introduces an innovative ensemble-based model designed to classify malware directly from binary file representations, bypassing the constraints of conventional approaches. The proposed method demonstrates a substantial improvement in accuracy, outperforming established machine learning and deep learning frameworks like VGG16, CNN, and XGBoost by achieving a remarkable accuracy of 98.75%. To optimize classification results further, the study incorporates advanced ensemble strategies, including a Voting Classifier, Decision Tree, and the Xception model, with the potential to achieve 99.85% accuracy. Furthermore, a user-friendly Flask-based web interface with secure authentication mechanisms was developed, enabling convenient and protected access to the classification system. This comprehensive framework delivers a compelling, accurate, and practical solution for malware classification, contributing significantly to enhancing cybersecurity systems.

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

Machine Learning Based Ensemble Learning Approach for Malware Family Classification Using Visual File Representations

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dubariya Vaishali,
  • L. Arokia Jesu Prabhu

摘要

The fast evolution of new malware variants poses a formidable challenge to cybersecurity, underscoring the critical necessity for sophisticated and robust classification techniques. Traditional techniques, which often depend on analyzing disassembled code, face limitations such as susceptibility to anti-disassembly strategies and increased error rates, reducing their efficacy in malware detection. To overcome these challenges, the proposed work introduces an innovative ensemble-based model designed to classify malware directly from binary file representations, bypassing the constraints of conventional approaches. The proposed method demonstrates a substantial improvement in accuracy, outperforming established machine learning and deep learning frameworks like VGG16, CNN, and XGBoost by achieving a remarkable accuracy of 98.75%. To optimize classification results further, the study incorporates advanced ensemble strategies, including a Voting Classifier, Decision Tree, and the Xception model, with the potential to achieve 99.85% accuracy. Furthermore, a user-friendly Flask-based web interface with secure authentication mechanisms was developed, enabling convenient and protected access to the classification system. This comprehensive framework delivers a compelling, accurate, and practical solution for malware classification, contributing significantly to enhancing cybersecurity systems.