Collective Anomaly Detection Using Big Data Distributed Stream Analytics

The majority of today's data is creating in a form of streams by diverse application domains and the sizes of these data are becoming very large in order to detect unusual events or predict intelligent decision. This often poses a significant challenging task to process and discover unusual events from such dynamic behaviours. Importantly, real time detection in streaming application is a critical concern to detect anomalous event, this is due to a potential infinite size of the data and an evolving behaviour of the event stream. In this paper, we proposed a novel distributed real time Collective Anomaly detection by; a) defining event stream snapshot model, b) designing a novel distributed Collective event stream detection, c) evaluates the experimental result in terms of the accuracy of the detection rates and the detection processing runtime performance. The experimental evaluation is demonstrated that our method is capable to detect a collection of events from high volumes of unbounded streams in real-time.