AI Method Developed to Predict Silent Internet Outages Before User Impact
Researchers from Istanbul Technical University (ITU) have developed an AI-based method designed to identify early warning signs of 'black holes' in internet networks. These 'black holes' are silent failures where routers appear to function normally but drop data packets without notification, leading to connection issues or data loss for users. Unlike traditional methods that react to packet loss, this new approach aims to predict these failures by detecting subtle, unusual behaviors in network traffic that are difficult to distinguish from normal activity. The study, published in IEEE Transactions on Network and Service Management, integrates various machine-learning techniques to monitor how backbone network traffic evolves over time. The system learns from unlabeled network traffic, identifying abnormalities and analyzing sequences to find subtle warning signs. This three-stage forecasting model combines clustering, time-series analysis, and deep learning.
