Protective Security Resource Allocation Using Stackelberg Security Games and Reinforcement Learning

Eligibility: UK/International (including EU) graduates with the required entry requirements

Duration: Full-time – between three and three and a half years fixed term

Application deadline: 25 Aug 2026

Interview date: Will be confirmed to shortlisted candidates

Start date: January 2027

For further details contact: Professor Andrew Glazzard

Introduction

Allocating protective security resources across sites and locations in the UK and South Africa is made continuously and under significant uncertainty. Security managers at transport hubs, event venues, government buildings, and critical national infrastructure (CNI) sites must decide where to place personnel, how to configure access controls, and when to heighten or relax security postures. In the UK, the Terrorism (Protection of Premises) Act 2025 (‘Martyn’s Law’) obliges those responsible for the largest capacity sites to provide counter-terrorist risk management responses. UK police deploy assets to public spaces to detect and deter hostile activity and provide advice to security managers in the private sector and in local authorities according to their estimates of where risk is greatest.

The protective security resource allocation problem suggests an intellectually stimulating challenge: which approaches in computer science and mathematical modelling might be applied to represent or simulate the behaviour of adversaries in choosing targets and responding to real and perceived security measures? Game theory provides one potential approach: security games model the interaction between attackers and defenders, who develop initial allocation strategies and respond dynamically to their adversaries’ strategies. A potentially complementary approach is reinforcement learning (RL), which provides algorithms capable of learning good policies in large, complex environments where analytical solutions are intractable.

Project details

This project is timely because protective security decision-making is moving into a more regulated, accountable and data-informed environment. Martyn’s Law increases the need for auditable protective security risk management, while police and security managers continue to face resource constraints across heterogeneous sites. The strategic opportunity is to develop decision-support methods that are explainable, empirically grounded and robust to adaptive adversary behaviour.

Despite the maturity of security game theory and RL, their application to protective security contexts is almost entirely absent from the published literature. This PhD will address this gap. Its proposed research questions are:

RQ1: How can protective security allocation across heterogeneous sites be formally specified as a defender–attacker resource allocation problem?
RQ2: How do allocation policies differ under progressively richer adversary models: perfect best response, bounded rationality, data-driven probabilistic response, and adaptive RL?
RQ3: Under what conditions do RL-based approaches outperform classical security-game approaches in terms of expected loss, robustness, scalability and adaptability?
RQ4: How can improved site-risk estimates be integrated into the allocation model?
RQ5: How do model-generated allocation policies compare with current protective security practice in practitioner-facing evaluation exercises?

The project will use a mixed methodology combining real-world problem specification, simulation-based model development, and practitioner validation. Because protective security resource allocation cannot be experimentally manipulated in the real world, the core comparative work will take place in simulation. However, the simulation will be grounded in real-world site typologies, resource constraints, open-source contextual data, and practitioner input from protective security stakeholders. The simulation is not intended to be a substitute for real-world validation; rather, it provides a controlled environment in which different allocation methods can be compared systematically before their assumptions and outputs are tested with practitioners.

This PhD is a cotutelle PhD between Coventry University (UK) and Stellenbosch University (South Africa). The successful candidate will primarily be based in Coventry, but will be required to undertake a significant period of study in South Africa as a requirement of the PhD.

Funding

Tuition fees and bursary.

Benefits

The successful candidate will receive comprehensive research training including technical, personal and professional skills. All researchers at Coventry University (from PhD to Professor) are part of the Doctoral and Researcher College, which provides support with high-quality training and career development activities.

Entry requirements

A minimum of a 2:1 first degree in a relevant discipline/subject area with a minimum 60% mark in the project element or equivalent with a minimum 60% overall module average.

PLUS

A master's degree in a relevant subject area.The master's must have been attained with minimum overall marks at merit level (60%)*. In addition, the dissertation or equivalent element in the master's must also have been attained with a minimum mark of merit level (60%).
The potential to engage in innovative research and to complete the PhD within 3.5 years.
A minimum of English language proficiency (IELTS academic overall minimum score of 7.0 with a minimum of 6.5 in each component).

Essential skills and knowledge

A first-class or upper second-class honours degree (or equivalent) in Computer Science, Artificial Intelligence, Data Science, Operations Research, Mathematics, Statistics, Engineering, or a closely related quantitative discipline.
Strong analytical and problem-solving skills, with the ability to formulate and solve complex technical problems.
Good programming skills in at least one language such as Python, Java, Julia, or C++.
A solid mathematical background, including probability, optimisation, linear algebra, and statistics.
An interest in artificial intelligence, decision support, optimisation, or computational modelling.
Ability to work independently while contributing effectively within a multidisciplinary research team.
Excellent written and verbal communication skills in English.
Strong organisational skills and the ability to manage a long-term research project.

Desirable knowledge and experience

Knowledge of optimisation techniques, mathematical programming, or operations research.
Familiarity with game theory.
Experience with machine learning, probabilistic modelling, or reinforcement learning.
Experience using optimisation or AI libraries (e.g., PyTorch, TensorFlow, Stable-Baselines3, OR-Tools, Gurobi, CPLEX, or similar).
Experience analysing and visualising data using Python or R.
Knowledge of simulation, agent-based modelling, or stochastic modelling.
Experience conducting empirical evaluations or experimental studies.
Familiarity with cyber security, protective security, critical infrastructure, or risk assessment (beneficial but not essential).
Evidence of scientific writing, such as a dissertation, publication, or technical report.

Personal attributes

Curious, creative, and motivated to tackle challenging interdisciplinary research questions.
Willingness to learn new mathematical, computational, and AI techniques.
Ability to critically evaluate competing methods and communicate findings clearly.
Resilient and persistent when working on open-ended research problems.
Enthusiastic about collaborating with academic and industry stakeholders.
Committed to conducting rigorous, reproducible, and ethical research.

How to apply

To find out more about the project, please contact Professor Andrew Glazzard.

All applications require full supporting documentation, a covering letter, plus a 2000-word supporting statement showing how the applicant’s expertise and interests are relevant to the project.

Apply to Coventry University