Scholarlink

Using machine learning, the study leverages real-world RIPE Atlas measurements to predict DNS latency spikes. Approximately 700 K ICMP-ping samples were collected over six months from 14 probes in the Kingdom of Bahrain. We evaluated four models: Random Forest, Decision Tree, Logistic Regression, and an LSTM deep learning, using engineered features such as average/min/max latency, jitter, and time-of-day indicators. Random Forest achieved 96% recall on the spike class and 99% overall accuracy, outperforming the other models. The results demonstrate that a feature-rich, time-series approach can proactively warn of DNS-latency spikes in the Kingdom of Bahrain and comparable environments.

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

Machine Learning-Based Prediction of DNS Latency Spikes Using RIPE Atlas: A Study from the Kingdom of Bahrain’s Perspective

  • 章节

  • 发表日期：

    2026年

查看全文
  • Sameh Foulad,
  • Sami Dagash

摘要

Using machine learning, the study leverages real-world RIPE Atlas measurements to predict DNS latency spikes. Approximately 700 K ICMP-ping samples were collected over six months from 14 probes in the Kingdom of Bahrain. We evaluated four models: Random Forest, Decision Tree, Logistic Regression, and an LSTM deep learning, using engineered features such as average/min/max latency, jitter, and time-of-day indicators. Random Forest achieved 96% recall on the spike class and 99% overall accuracy, outperforming the other models. The results demonstrate that a feature-rich, time-series approach can proactively warn of DNS-latency spikes in the Kingdom of Bahrain and comparable environments.