Technical note | Future Proofing Computational Red Teaming
Computational Red Teaming describes the application of new and innovative analytic techniques, tools and methodologies in support of Red Teaming Activities. This report documents the formal design of such a tool, to inform the future direction of the DSTO research program. Requirements for an executable prototype are presented to construct the initial design, developed through the application of Systems Engineering practices. Recommendations to further develop the prototype are provided, with illustrative examples. When implemented, the prototype will deliver a Computational Red Teaming capability to assess technologies and systems, concepts and force structures; with the potential to inform better decisions in future Australian capability development.
Joint Operations Division established an initiative in Computational Red Teaming in 2010, under the Defence Science and Technology Organisation's Corporate Enabling Research Program (CERP). The program conducts fundamental research in the field, evaluates new and innovative analytic techniques, develops computational support tools and applies the methodologies of Red Teaming in support of activities and experimental campaigns.
This study follows from, and implements, the recommendations of two earlier scoping studies, delivered under the Computational Red Teaming initiative by Gowlett (2011) and Wheeler (2012). They recommend the development of a tool for the purposes of assessing Australian capability, concepts and force structures. Additional guidance for the CERP identifies the preferred domain of application for the tool should be in support of study and analysis for countering Improvised Explosive Devices.
Our report documents the start of the formal design process for a Computational Red Teaming Tool, in line with the existing program of work, and to inform the future direction of the corporate research program. Conceptual requirements for an executable prototype are presented and a design developed through the application of Systems Engineering practices. Illustrative examples for possible implementations of the design are also provided together with an example case study for application to the CIED domain.
The proposed prototype is to demonstrate the proof-of-concept. Further design work is required. However, should the prototype be successful, a wider program of research could be initiated. If implemented, the prototype will deliver a Computational Red Teaming capability to assess technologies and systems, concepts and force structures; with the potential to inform better decisions in future Australian capability development.
The following recommendation is advised, for forward work planning within the Joint Operations Division's Computational Red Teaming Task.
The Computational Red Teaming Program continues to develop the prototype design presented in this report. Progression of this design to the next stage requires:
selection of a case study or problem to be addressed in the CIED domain;
identification of stakeholders, clients and end-users;
development of use-cases, outlining how those end-users are to interact with the software; and
trials of those interactions through example Graphical User Interfaces (GUIs).
Should this recommendation be supported, a team of at least four major skill sets will be required. These include: software design; systems engineering; human-factors; and program management. The team would take the initial conceptual design presented in this document and develop the customer requirements specification. This requirements specification will guide the development of this concept into the prototype.
Finally, the purpose of this report is to deliver a design for a Computational Red Teaming software tool. To this end, the science of adversarial reasoning component has been less important than the strength of the design. We then present one further recommendation.
The Computational Red Teaming Program initiate a study to proceed this work. The study will report on follow-on options to implement the adversarial reasoning and logic components of the design. In doing so, the relevant techniques from the discipline of Machine Learning and Artificial Intelligence will be considered. This study will be conducted in parallel with Recommendation 1 and its outcomes will inform the implementation of the design.