Syslog Anomaly Detection Using Supervised Machine Learning Models

Abstract

Nowadays, detecting anomalies is crucial for managing every network. Massive logs are produced by modern large-scale distributed systems. These logs contain useful information regarding network behavior. Traditionally, developers detect anomalies by complex coded scripts. However, such approach is not efficient for large-scale systems where they generate thousands of logs. Thus, syslog anomaly detection tool has been proposed in this paper by using supervised machine learning (ML) models. As a source of dataset for the ML models, syslog generator was developed to generate the desired dataset. A comparative study about many supervised ML methods has been evaluated in this paper using different amount of datasets. The target was to check the impact of enlargement of datasets on the performance of the anomaly detections.