Scholarlink

<p>Mobile web applications support important services like banking, e-commerce, healthcare as well as communication in the current digital era. However due to their widespread usage they are now more vulnerable to complex cyberattacks. The dynamic and resource-constrained characteristic of mobile settings is frequently too much for conventional safety protections to handle. In order to effectively identify vulnerabilities and intrusions in mobile online systems, this study suggests a hybrid cybersecurity framework that combines Ant Colony Optimisation (ACO) with Randomised Decision Tree Classifier (RDTC). Unlike existing optimisation–classifier hybrids the proposed ACO-RDTC introduces an adaptive feature–subset refinement mechanism specifically tailored for mobile web environments enabling efficient handling of dynamic traffic patterns and resource limitations. The approach employs comprehensive data preprocessing using Singular Value Decomposition (SVD) for dimensionality reduction as well as SelectKBest for appropriate feature selection. This combination was chosen because RDTC effectively manages heterogeneous decision boundaries, ACO offers excellent global optimisation capabilities as well as SVD with SelectKBest aids in noise reduction while preserving crucial discriminative features. detection and cyber threats. ACO further improves RDTC performance by dynamically optimizing hyperparameters and selecting the most discriminative feature subsets. Using the CSE-CIC-IDS 2018 dataset, the proposed ACO-RDTC model achieved 99.08% accuracy, 98.50% precision, 99.08% recall and 98.74% F1-score, outperforming existing tactics as well as considerably reducing false positives. Furthermore an ablation study validated the contribution of each module (SVD, SelectKBest and ACO), while cross-dataset testing on UNSW-NB15 recognised the model’s generalizability with 97.82% accuracy, emphasising its robustness as well as scalability for securing modern mobile web 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

ACO-tuned randomized decision tree for detecting cyber threats in mobile web applications

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月19日

查看全文
  • Abeer Aljohani

摘要

Mobile web applications support important services like banking, e-commerce, healthcare as well as communication in the current digital era. However due to their widespread usage they are now more vulnerable to complex cyberattacks. The dynamic and resource-constrained characteristic of mobile settings is frequently too much for conventional safety protections to handle. In order to effectively identify vulnerabilities and intrusions in mobile online systems, this study suggests a hybrid cybersecurity framework that combines Ant Colony Optimisation (ACO) with Randomised Decision Tree Classifier (RDTC). Unlike existing optimisation–classifier hybrids the proposed ACO-RDTC introduces an adaptive feature–subset refinement mechanism specifically tailored for mobile web environments enabling efficient handling of dynamic traffic patterns and resource limitations. The approach employs comprehensive data preprocessing using Singular Value Decomposition (SVD) for dimensionality reduction as well as SelectKBest for appropriate feature selection. This combination was chosen because RDTC effectively manages heterogeneous decision boundaries, ACO offers excellent global optimisation capabilities as well as SVD with SelectKBest aids in noise reduction while preserving crucial discriminative features. detection and cyber threats. ACO further improves RDTC performance by dynamically optimizing hyperparameters and selecting the most discriminative feature subsets. Using the CSE-CIC-IDS 2018 dataset, the proposed ACO-RDTC model achieved 99.08% accuracy, 98.50% precision, 99.08% recall and 98.74% F1-score, outperforming existing tactics as well as considerably reducing false positives. Furthermore an ablation study validated the contribution of each module (SVD, SelectKBest and ACO), while cross-dataset testing on UNSW-NB15 recognised the model’s generalizability with 97.82% accuracy, emphasising its robustness as well as scalability for securing modern mobile web environments.