Computational red teaming
Red teaming is a method for assessing vulnerabilities in systems or structures. Two factions or teams – red and blue – are posited or formed. The red team is charged with attacking the system or structure being defended by the blue team. The role of the red team is to challenge the implicit assumptions in blue team defences.
Computational red teaming (CRT) uses computational models to enhance red teaming. Using CRT, many scenarios can be simulated and analysed (whereas manual red teaming can only manage a handful). This makes it possible to gather enough data to thoroughly understand the situation being modeled, using statistical analysis, search-based optimisation and machine learning methods. These methods can often uncover unexpected knowledge, as computerised methods do not suffer from ‘blind spots’ in the same way that humans do.
This project will develop a CRT platform based on agent-based simulation, evolution-based optimisation and cluster/cloud computing, and will use this platform to investigate red teaming applications in various domains including military (combat scenarios and other) and civilian (for example critical infrastructure protection) applications.
The project will build on existing seed work with DSO Laboratories and Nanyang University in Singapore, Dortmund Technical University in Germany, and DSTO Joint Operations Division.