Scholarlink

This study aims to address the critical challenge of class imbalance in cyberattack data using advanced deep learning techniques. Imbalance in datasets can lead to biased model predictions, disproportionately favoring the majority class while underperforming on minority attack types. To counter this, we applied Bayesian Neural Networks (BNN), Naive Bayes, and Long Short-Term Memory (LSTM) models, achieving an accuracy of 88% with the LSTM model, the highest out of 10 iterations. Both BNN and LSTM demonstrated their efficacy in cyberattack classification. To mitigate the effects of class imbalance, we applied mathematical sampling techniques, ensuring a more equitable representation of classes. The Kolmogorov-Smirnov (KS) test was employed to evaluate the models’ performance and understand the data distribution. Visual representations of each class versus others, with p-values from the KS test, were used to highlight distributional disparities, supporting the statistical analysis and enhancing dataset understanding.

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

Deep Learning and Statistical Testing Approaches for Cyberattack Classification in Imbalanced Datasets

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • G. Ishvarya,
  • Kavya Sree Kammari,
  • Sharanya Vanraj Thambi,
  • Sarada Jayan

摘要

This study aims to address the critical challenge of class imbalance in cyberattack data using advanced deep learning techniques. Imbalance in datasets can lead to biased model predictions, disproportionately favoring the majority class while underperforming on minority attack types. To counter this, we applied Bayesian Neural Networks (BNN), Naive Bayes, and Long Short-Term Memory (LSTM) models, achieving an accuracy of 88% with the LSTM model, the highest out of 10 iterations. Both BNN and LSTM demonstrated their efficacy in cyberattack classification. To mitigate the effects of class imbalance, we applied mathematical sampling techniques, ensuring a more equitable representation of classes. The Kolmogorov-Smirnov (KS) test was employed to evaluate the models’ performance and understand the data distribution. Visual representations of each class versus others, with p-values from the KS test, were used to highlight distributional disparities, supporting the statistical analysis and enhancing dataset understanding.