Activity-Attack Graphs for Intelligence-Informed Threat COA Development

Abstract

A threat course of action (COA) describes the likely tactics, techniques, and procedures (TTPs) an adversary may deploy across the cyber kill-chain. Threat COA development and analysis informs hunt teams, incident responders, and threat emulation efforts on likely activities the adversary will conduct during an attack. In this paper, we propose a novel approach to generate and evaluate threat COAs through association rule mining. We identify frequent TTP itemsets to create a set of activity groups that describe associations between TTPs. We overlay activity groups to create a directed and edge-weighted activity-attack graph. The graphs hypothesize various adversary avenues of attack, and the weighted edges inform the analyst's trust of a hypothesized TTP in the COA. Our research identifies meaningful associations between TTPs and provides an analytical approach to generating threat COAs. Further, our implementation uses the STIX framework for extensibility and usability in a variety of threat intelligence environments.