Scholarlink

<p>Mare exploits systems intending to disrupt operations or compromise sensitive data, and it represents an extensive and dynamic threat landscape. Effectiveness across platforalwms, generalization, and adaptability are frequently challenges for traditional malware detection techniques, particularly when dealing with emerging malware varieties. To overcome these challenges, this paper proposes a Hybrid Stacking-based Ensemble Framework for Malware Detection (HSEF-MD) that covers three significant gaps in prior research: limited integration of dynamic analysis, insufficient cross-platform feature integration, and insufficient training data. HSEF-MD builds a strong training basis through acquiring and combining sizable static and dynamic datasets from the Windows and Android platforms. HSEF-MD aims to enhance overall accuracy by addressing concept drift, minimizing data imbalance, improving model interpretability, and reducing the false positive rate (FPR). A number of research studies, comprising three different trials, show that incorporating dynamic analysis into the training phase considerably improves model performance. Evaluation measures, including F1 score, ROC-AUC, accuracy, precision, and FPR, confirm the efficiency and supremacy of the HSEF-MD over conventional models. Our findings highlight the importance of dynamic feature integration and cross-platform fusion are to building robust malware detection systems. Notably, the HSEF-MD demonstrates robustness in cross-platform scenarios, such as training on dynamic features and validating on static features, offering practical cost advantages with minimal performance compromise.</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

Feature Fusion-Based Ensemble Approach for Robust Malware Detection with Reduced False Positives

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Asia Maqsood,
  • Hamid Turab Mirza,
  • Faiza Iqbal,
  • Reem Ibrahim Alkanhel,
  • Naveed Hussain,
  • Ayesha Afzaal,
  • Ayesha Altaf,
  • Imran Ashraf

摘要

Mare exploits systems intending to disrupt operations or compromise sensitive data, and it represents an extensive and dynamic threat landscape. Effectiveness across platforalwms, generalization, and adaptability are frequently challenges for traditional malware detection techniques, particularly when dealing with emerging malware varieties. To overcome these challenges, this paper proposes a Hybrid Stacking-based Ensemble Framework for Malware Detection (HSEF-MD) that covers three significant gaps in prior research: limited integration of dynamic analysis, insufficient cross-platform feature integration, and insufficient training data. HSEF-MD builds a strong training basis through acquiring and combining sizable static and dynamic datasets from the Windows and Android platforms. HSEF-MD aims to enhance overall accuracy by addressing concept drift, minimizing data imbalance, improving model interpretability, and reducing the false positive rate (FPR). A number of research studies, comprising three different trials, show that incorporating dynamic analysis into the training phase considerably improves model performance. Evaluation measures, including F1 score, ROC-AUC, accuracy, precision, and FPR, confirm the efficiency and supremacy of the HSEF-MD over conventional models. Our findings highlight the importance of dynamic feature integration and cross-platform fusion are to building robust malware detection systems. Notably, the HSEF-MD demonstrates robustness in cross-platform scenarios, such as training on dynamic features and validating on static features, offering practical cost advantages with minimal performance compromise.