Scholarlink

<p>Ransomware attacks, which encrypt a victim’s data and demand cryptocurrency payment for decryption, pose serious operational and financial threats to individuals, organizations, and governments. As these attacks become increasingly prevalent, there is a critical need for predictive models that can proactively mitigate their impact. This study proposes a novel framework combining machine learning (ML) techniques, Chi-Square feature selection, and hyperparameter optimization via the African Vultures Optimization Algorithm (AVOA) to enhance ransomware detection. Three ML models Adaptive-Boost Learning Classifier (ADAC), Extreme Learning Machine (ELM), and Extra Trees Classifier (ETC) were integrated with AVOA to improve classification performance. The Chi-Square method identified Financial Impact (USD) and Clusters as the most influential features, significantly boosting model accuracy. Among the evaluated models, the Extra Trees Classifier optimized with AVOA (ETAV) demonstrated superior performance, achieving a test accuracy of 0.981 and maintaining high precision across all severity levels (Severe, Significant, Average). The proposed framework offers a robust and interpretable approach for ransomware attack prediction, supporting more informed cybersecurity decision-making.</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

Generative mathematical models for ransomware attack prediction using Chi-Square feature selection for enhanced accuracy

  • 期刊论文

  • 发表日期：

    2025年6月26日

查看全文
  • Ruofan Su

摘要

Ransomware attacks, which encrypt a victim’s data and demand cryptocurrency payment for decryption, pose serious operational and financial threats to individuals, organizations, and governments. As these attacks become increasingly prevalent, there is a critical need for predictive models that can proactively mitigate their impact. This study proposes a novel framework combining machine learning (ML) techniques, Chi-Square feature selection, and hyperparameter optimization via the African Vultures Optimization Algorithm (AVOA) to enhance ransomware detection. Three ML models Adaptive-Boost Learning Classifier (ADAC), Extreme Learning Machine (ELM), and Extra Trees Classifier (ETC) were integrated with AVOA to improve classification performance. The Chi-Square method identified Financial Impact (USD) and Clusters as the most influential features, significantly boosting model accuracy. Among the evaluated models, the Extra Trees Classifier optimized with AVOA (ETAV) demonstrated superior performance, achieving a test accuracy of 0.981 and maintaining high precision across all severity levels (Severe, Significant, Average). The proposed framework offers a robust and interpretable approach for ransomware attack prediction, supporting more informed cybersecurity decision-making.