Simon Dietz, Kai Klede, An Nguyen, Bjoern M Eskofier
ContraLog is a parser-free method for log anomaly detection using contrastive learning and masked language modeling to predict message embeddings, showing effectiveness on benchmark datasets.
Logs are essential for understanding how computer systems are functioning, but detecting anomalies in these logs can be challenging. Traditional methods often simplify logs into templates, losing important details. ContraLog is a new approach that avoids this by using advanced techniques to analyze logs without pre-processing them into templates. It uses machine learning to understand the context and sequence of log messages, making it better at spotting unusual patterns or anomalies. The approach has been tested on several datasets and has shown promising results, offering a new way to detect issues in system logs.
