LogLAB: Attention-Based Labeling of Log Data Anomalies via Weak Supervision

With increasing scale and complexity of cloud operations, automated detection anomalies in monitoring data such as logs will be an essential part managing future IT infrastructures. However, many methods based on artificial intelligence, supervised deep learning models, require large amounts labeled training to perform well. In practice, this is rarely available because labeling log expensive, time-consuming, requires a understanding the underlying system. We present LogLAB, novel modeling approach for messages without requiring manual work by experts. Our method relies estimated failure time windows provided systems produce precise datasets retrospect. It attention mechanism uses custom objective function weak supervision techniques that accounts imbalanced data. evaluation shows LogLAB consistently outperforms nine benchmark approaches across three different maintains F1-score more than 0.98 even at windows.