Scholarlink

The prevalence of IoT botnet attacks has steadily increased as a result of the widespread use of IoT devices. Hence, it is crucial to identify botnet assaults in IoT network at an early stage. However, a steady rise in data packets causes an overfitting issue, and employing current models with inconsistent and ambiguous information about traffic patterns lowers the detection rate. To overcome this issue, in this paper a robust Fuzzy Chaotic Cuckoo Search Relief Feature selection algorithm (FCCRF) is developed. FCCRF handles inconsistent data by adopting fuzzy triangular membership function, chaotic mapping for searching a diverse population, and cuckoo searching strategy to find best feature subset most relevant to detect botnet attacks. This work used UNSW-NB 15 dataset and for validating support vector machine, naïve Bayes classifier and Logistic regression are applied on both reduced feature subset and whole feature set. The results proved feature subset of FCCRF improves the performance of SVM, NB, and LR classifiers with the accuracy of 0.92, 0.84, and 0.78%. And it is also observed that SVM produced highest detection rate in prediction of botnet attacks in IoT.

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

Robust Fuzzy Chaotic Cuckoo Search Boosted Relief Feature Selection for Dimensionality Reduction to Improve Botnet Attack Detection in IoT

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Veena Antony,
  • N. Thangarasu

摘要

The prevalence of IoT botnet attacks has steadily increased as a result of the widespread use of IoT devices. Hence, it is crucial to identify botnet assaults in IoT network at an early stage. However, a steady rise in data packets causes an overfitting issue, and employing current models with inconsistent and ambiguous information about traffic patterns lowers the detection rate. To overcome this issue, in this paper a robust Fuzzy Chaotic Cuckoo Search Relief Feature selection algorithm (FCCRF) is developed. FCCRF handles inconsistent data by adopting fuzzy triangular membership function, chaotic mapping for searching a diverse population, and cuckoo searching strategy to find best feature subset most relevant to detect botnet attacks. This work used UNSW-NB 15 dataset and for validating support vector machine, naïve Bayes classifier and Logistic regression are applied on both reduced feature subset and whole feature set. The results proved feature subset of FCCRF improves the performance of SVM, NB, and LR classifiers with the accuracy of 0.92, 0.84, and 0.78%. And it is also observed that SVM produced highest detection rate in prediction of botnet attacks in IoT.