Information fusion approach for anomaly detection in big data

点击翻译按钮获取中文摘要

The recent focus on big data analytics to improve situational awareness and decision making have sparked an interest in autonomous processing of sensory data. The ability to detect anomalies and identify abnormal situations can empower organizations to make informed decisions and ultimately reduce costs. Anomaly detection is a complex problem, but using big data and information from multiple, non-homogeneous sources greatly compounds the difficulty. Not only do anomalies change with time, but early detection requires a system to learn continuously. Current methodologies cannot be used efficiently to detect anomalies. The objective of the proposed project is to develop an information fusion-based anomaly detection system, utilizing the Observe, Orient, Decide, Act model of human decision making for situational awareness. The anomaly detection system will be capable of autonomously processing multiple, continuous sensing information in a dynamic environment. We will develop and test a software-based implementation on two applications: airborne sensing and pipeline monitoring. The anticipated research results on anomalous patterns will provide insight to situational anomalous events. This will enable organizations to make informed decisions in an evolving world of larger data. Another direct benefit is the potential for the results to be integrated in existing products or patents with our industrial partners. This project will be of great impact for next generation anomaly detection technologies to be deployed in the industry, which the high demand for faster, real-time monitoring and analysis is of major importance.