Incremental Theory Closure Reasoning for Large Scale Knowledge Graphs

Abstract

Knowledge graph(KG)s have become more and more important for the field of cybersecurity. However, they usually contain much implicit semantic information, which needs to be further mined through semantic inference to be more useful for both analysts and automated systems. In this paper, we propose an incremental reasoning algorithm KGRL-Incre for the scenario that the instances of a KG is expanded with only a small set of triples, which can perform an incremental update to the previous reasoning result effectively to avoid a full re-reasoning over the expanded KG. The main contributions of our approach are the irrelevant triple filtering algorithms which reduce the scale of data that need to be processed and a delay reasoning strategy which limits the number of time-consuming iterations while still preserves relative completeness of the final result. The extensive experiments and comprehensive evaluations are conducted and the experimental results show that the KGRL-Incre can significantly reduce time consumption compared with the expanding and reasoning approach in the target scenario.