A Survey on Anomaly Detection Methods for System Log Data

System logs are often a collection of unrelated print statements which records certain events that occur while the system is running. Log file analysis often can be crucial for finding system faults which can otherwise be quite difficult to detect. Traditional log analysis involves analyzing line by line until a discrepancy is spotted. This process is tedious, time consuming and is prone to human errors. With the advent of machine learning, several new methods have been devised which can make anomaly detection much easier. Further, in the past decade, deep learning has evolved so much that new techniques and algorithms spring every now and then. This paper examines several existing techniques that can be used for system log analysis.