Automatic Summarization of Validated Intelligence Events

Abstract

In recent years there have been enormous progress and breakthroughs in the field of natural language processing (NLP). These breakthroughs have significantly advanced the state-of-the-art in NLP across the board and there have been a growing interest to apply these findings in an industrial setting. This thesis work is carried out in collaboration with Recorded Future, that has an interest in whether large language models can produce Validated Intelligence Event (VIE) summaries of high quality. A VIE summary is an analytical offering that describes an event in the cybersecurity domain and if these could be automatically generated it would allow for higher throughput of such summaries and would generate more value for their clients. Our results show that such summaries are possible to produce with relatively high performance, even though they cannot be completely automated with the techniques used in this paper. However, the analysts who are currently producing these summaries expect that the use of an automated system such as this can decrease the production time by 4.