A Machine Learning approach to Generic Entity Resolution in support of Cyber Situation Awareness

Moir, C. and Dean, J.

This paper introduces the Generic Entity Resolution (GER) framework; a framework that classifies pairs of entities as matching or non-matching based on the entities’ features and their semantic relationships with other entities. The GER framework has been developed as part of an AI-based system for the development of Cyber situational awareness and provides a data fusion role by resolving entities discovered across multiple disparate data sources. The approach utilizes supervised machine learning to identify the set of features and semantic relationships that result in the optimum classification accuracy. We evaluated the GER framework using several well-known data sets and compare the framework’s accuracy to existing state-of-the- art resolution algorithms. We found that the GER framework’s accuracy compares favourably to existing state-of-the-art resolution algorithms for the data sets used in this evaluation.

Cite as: Moir, C. and Dean, J. (2015). A Machine Learning approach to Generic Entity Resolution in support of Cyber Situation Awareness. In Proc. 38th Australasian Computer Science Conference (ACSC 2015) Sydney, Australia. CRPIT, 159. Parry, D. Eds., ACS. 47-58

(from crpit.com) (local if available)