Introduction
The defence and security (D&S) community – represented by the Ministry of Defence (Defence Science and Technology Laboratory [Dstl] and Joint Forces Command), GCHQ and MI5 – are collaborating with The Alan Turing Institute to deliver an ambitious programme of data science and artificial intelligence (AI) research to deliver impact in real world scenarios.
Following the signing of a collaboration agreement and a year of knowledge exchange the D&S programme was launched in 2017. Since then, programme personnel has grown across research, leadership, and support.
The programme’s vision is to undertake multidisciplinary data science and AI research to ensure a safe, secure, and prosperous society:
- Safe – supporting defense and national security agencies to keep societies and citizens safe
- Secure – protecting the privacy and security of citizens, institutions and industry
- Prosperous – contributing to global good by enabling societies around the world to derive benefit from strategy technological advances
The programme carries out this vision through traditional academic research (e.g. DSO Partnership, AI for Cyber Defence Research Centre [AICD]), capability development (e.g. Applied Research Centre for Defence and Security [ARC]) and policy analysis (e.g. Centre for Emerging Technology and Security [CETaS])
Programme challenges
The D&S programme is solving challenges across four key areas:
Cyber, privacy, trust and identity
Existing projects:
- Developing novel solutions for human-machine collaboration
and autonomous defence against a range of cyber incursions;
- testing and building capabilities to generate synthetic datasets;
- releasing new secure machine learning models and protocols (including new anonymisation techniques).
Decision support
Existing projects:
- Improving the representation of uncertainty for decision makers;
- the use of complex networks and spatial interaction theory to model the effect of regional global interactions;
- benchmarking the efficacy of using reinforcement learning techniques for wargaming.
Global good
Existing projects:
- Establishing a deeper understanding on how AI ethics can be applied across a range of D&S contexts;
- providing a detailed background on the risks that climate change poses to UK security;
- understanding the spatial and temporal processes of conflict, climate change and migration, and how these lead to potential risks of exploitation.
Understanding data
Existing projects:
- Better detection and classification of patterns and anomalies within datasets of cyber defence relevance;
- the development of Bayesian deep learning techniques;
- new topological data analysis techniques.
Impact
Starting with the identification of three core research projects, the programme’s activity has rapidly expanded. Additional research projects in our remaining two key challenge areas will begin before the end of the year, and several interdisciplinary research themes have evolved that cut across the Turing’s strategic partners. Our activities are driven by the programme’s goals of delivering world-leading research with real-world impact.
Developing state-of-the-art defences for computer networks
Every computer network is vulnerable to cyber criminals and hackers. Detecting attackers within a network is a complex task, and even skilled operators struggle to keep track of them all.
To bolster network defences and speed up response times, computer security organisations are increasingly looking to autonomous systems. Researchers led by the Turing’s Vasilios Mavroudis and Chris Hicks are exploring a technique called reinforcement learning – a field of artificial intelligence in which computer algorithms learn by solving problems through trial and error, with the goal of maximising a specified reward.
In 2022, the team won the first Cyber Autonomy Gym for Experimentation (CAGE) challenge, run by The Technical Cooperation Program. The team has made its code publicly available so that others can benefit from and expand upon this work.
Read the full case study as seen in the 2021-22 Alan Turing Institute Annual Report
Short-term projects aim to demonstrate immediate, meaningful impact
An announcement was made that the programme’s long-term work has been bolstered by a number of shorter, strategically important projects supported by funding from GCHQ. Each up to six months in duration, these projects aim to demonstrate immediate, meaningful impact, and address the key challenges that frame the defence and security programme.
The projects are focusing on a diverse range of applications including understanding hacker communities, adversarial machine learning, encryption, modelling of civil conflict, topological data analysis, and utilising game theory in cyber security. The projects are expected to yield academic impact through publications, and real-world impact through software, which will be released for use and further development.
The Manufacturer, the UK’s premier industry publication for providing manufacturing news, articles, and insights, published an article about the announcement.
Mapping conflict data to explain, predict, and prevent violence
Work produced by Turing Fellow Dr Weisi Guo is aiming to understand the mechanics that cause conflict and identify multi-scale population areas that are at risk of conflict. The research utilises the latest developments in complex networks and spatial interaction theory to model the effect of multiplexed regional-global interactions.
Findings have shown that ‘crossroad’ towns and cities where there are few other routes correlate strongly with data on violence; including terrorism, war between states, and gang violence. The work is building an evidence base which aims to help sustainable global development of infrastructure in order to reduce conflict.
Dr Guo was interviewed by BBC News about his work, the article going into depth about the potential ramifications of the work and the role of the Turing in the research.
Read the BBC article – “Can mapping conflict data explain, predict and prevent violence?”
Retool AI to forecast and limit wars
Turing Fellow Dr Weisi Guo and Director of Special Projects Sir Alan Wilson have written a comment piece for Nature, explaining how using artificial intelligence to predict outbursts of violence and probe their causes could save lives.
The piece details existing research being conducted into the forecasting of conflict, including the work of Guo and Wilson at the Turing. The piece identifies three things that will improve conflict forecasting: new machine-learning techniques; more information about the wider causes of conflicts and their resolution; and theoretical models that better reflect the complexity of social interactions and human decision-making. The piece goes on to propose that an international consortium be set up to develop formal methods to model the steps society takes to wage war.
Organisers
George Balston
Programme Co-Director, Defence and Security
Alaric Williams
Programme Manager, Defence and Security
Tracey Peterson
Senior Programme ManagerContact info
For more information, please contact Tracey Peterson ([email protected]) and Hushpreet Dhaliwal ([email protected]).
