Non-Pattern-Based Anomaly Detection in Time-Series

Anomaly detection across critical infrastructures is not only a key step towards detecting threats but also gives early warnings of the likelihood potential cyber-attacks, faults, or infrastructure failures. Owing to heterogeneity and complexity cybersecurity field, several anomaly algorithms have been suggested in recent past based on literature; however, there still exists little no research that points focuses Non-Pattern Detection (NP-AD) Time-Series at time writing this paper. Most existing approaches refer initial profiling, i.e., defining which behavior represented by series “normal”, whereas everything does meet criteria “normality” set as “abnormal” anomalous. Such definition reflect sophistication nature. Under different conditions, same may be Therefore, authors paper posit need for NP-AD toward showing relevance deviating conforming expected behaviors. (NP), context paper, illustrates non-conforming patterns technique with respect some characteristics while dynamically adapting changes. Based experiments conducted it has observed significant approach margins data streams used from perspective non-seasonal outliers, Numenta Benchmark (NAB) dataset SIEM SPLUNK machine learning toolkit. It authors’ opinion provides predicting futuristic anomalies diverse cyber, infrastructures, other complex settings.