Cohort 2025 Projects

 

Deborah Amanor 

Beyond End-of-Life: Security Assurance for Legacy Vehicles in Intelligent Transportation Systems

Modern vehicles are being designed to have longer lifetimes, embedded with connectivity and autonomous capabilities that enable them to interact with digital infrastructure. As these vehicles age, they will eventually outlive manufacturer support, while the infrastructure and security requirements around them continue to evolve.

This project investigates how legacy connected vehicles can remain trustworthy participants in evolving smart infrastructure when security updates are no longer available and isolating or replacing these vehicles is impractical. The focus is on investigating the factors that lead to security assurance degradation, developing methods to track and quantify this degradation over time, and exploring mechanisms that enable legacy vehicles to continue participating in intelligent transportation systems (ITS) without compromising the security of the wider system. The broader aim of this project is to contribute to sustainable security engineering by moving towards continuous security assurance throughout a vehicle's operational lifetime, even beyond manufacturer support.

Dr Sana Belguith (Bristol)

Joel Anyam 

From Recovery to Recoverability: Continual Assurance of Backup and Recovery Systems

Organisations increasingly require controls that are verified, tested, and designed to ensure resilience in the face of rising real-world compromise. Resilience matters because prevention alone no longer suffices; the question for many organisations is not if they will face an attack, but when, and how effectively they can recover. Among the mechanisms that contribute to resilience, backups are pivotal: the National Cyber Security Centre (NCSC) identifies them as an essential component of response and recovery.

My current research interests centre on the cyber resilience of digital infrastructures, and my PhD focuses on the continual assurance of backup and recovery systems. I take a full-system approach (beyond data) and an end-to-end approach to assuring these systems. Through this research, I aim to assess the recoverability of these systems. I hope to see a shift from uncertain recovery attempts during a cyberattack to confident, resilient strategies through recoverability indicators.

Professor Awais Rashid (Bristol)

Nikki Bellosillo

Beyond Detection: Responsibility Gaps at the Payment-Stage of UK-Reported Cyber-Enabled Fraud

Cyber-enabled fraud in which people are deceived into sending money themselves remains one of the most common form of fraud in the UK. While mandatory reimbursement has been in force since October 2024, reimbursement only settles who bears the loss. Tracing, recovery, investigation, and referral of cases remain dispersed across banks, payment bodies, regulators, and law enforcement. When a case stalls after it is reported, existing records cannot show whether no organisation had the duty, capacity or route to act (a true gap), or whether several actors held partial, misaligned duties but none owned the case (crowded non-ownership). This study will reconstruct how individual cases move through this payment-response stage, using public records such as Financial Ombudsman Service decisions. It will map each organisation’s responsibilities, model the chain using law and economics, and test the model through interviews and a workshop with practitioners and victim-support organisations. The aim is to enable regulators and institutions to distinguish the two failures, so that responsibility and liability can be effectively reallocated.

Professor Adam Joinson (Bath)

Dr Joanna Syrda (Bath)

Maria Grigore 

Epistemic trust in OSINT: Provenance and the challenge of AI-generated content

Under what conditions can epistemic trust in OSINT be maintained when AI-generated content disrupts existing methods of provenance
verification?

This project reframes the post-trust problem in epistemic terms by looking at the conditions necessary under which any source can warrant belief, regardless of how it was produced (AI-generated or otherwise). The goal is to produce a guidance through Delphi validation with high-stakes OSINT actors, providing a decision-process practitioners can apply under real time and resource constraints.

Professor Adam Joinson (Bath)

Dr Tim Hill (Bath)

Dylan Jones 

Measuring and Improving the Fidelity of LLM-Orchestrated Synthetic Cybersecurity Telemetry

Defensive machine-learning systems that exist to protect modern infrastructure need to be able
to learn from security telemetry, yet the logs, network flows, and the endpoint events they
depend on are exactly what organisations are unable to share for reasons of privacy, regulation,
and infrastructure exposure. Public benchmark datasets are dated and demonstrably defective,
and the synthetic alternatives that promise to be able to replace them are validated only through
appeals to plausible appearance. The scientific gap is therefore one of measurement as much
as generation: no validated link exists between any measured notion of telemetry fidelity and
the defensive utility that fidelity is supposed to deliver.

This project therefore aims to develop a multi-dimensional framework for measuring that
fidelity and builds an LLM-orchestrated enterprise sandbox in which agents act as users, administrators, and attackers so that genuine systems emit telemetry natively, and determines
experimentally which fidelity dimensions predict transfer to unseen data. The outcome closes
the loop between generation, measurement and utility, and releases a validated framework,
reusable sandbox tooling and privacy-tested datasets.

Dr Alma Oracevic (Bristol)

Demi McCollum

Beyond the Individual: Collective Harm from Aggregated Data in Humanitarian Financial Systems

Refugee data is increasingly aggregated across aid and financial systems, and analysed with AI. This PhD investigates how this can expose groups of people to harm through demographically identifiable information (DII), meaning data that can be used to classify or track groups by characteristics such as gender or ethnicity, even when each person's data is handled properly. I will combine interviews with practitioners, system architects and refugees with computational experiments that measure how much group-level information can be inferred. Drawing on a realisation-based approach to justice, it asks what privacy protection should realistically mean in this context.

Dr Partha Das Chowdhury (Bristol)

Damien McGeer

Examining the Degradation of Secure Decision Making in Actor-Driven Insider Threats

Actor-driven insider threats are those who are the subject of social engineering and/or recruitment efforts from a malicious outside party, an aspect of the insider threat problem which is under-researched despite its atypical nature. Current research posits that actor-driven behaviour differs from expected threat activity, but the latent factors and theories at play have yet to be explored in depth. To address this problem, this project aims to examine the development of risk-taking behaviour and perception over the lifetime of an attack - how it may vary and the factors which could contribute to this change.

Dr Lukasz Piwek (Bath)

Dr Emma Walker (Bath)

Charles Ment 

Deceptive Promotional Practices for Online Gambling Operations on Live Streaming Platforms

Gambling live streams remain prominent on live streaming platforms, such as Kick or Twitch. Streamers on these platforms often promote offshore online casinos by playing them on stream and linking to them in their titles, descriptions and chat. Furthermore, engaging moments from these streams circulate across other social media platforms. The methods by which these casinos are promoted can often mislead the audience through deceptive framings of gambling activities or complex financial inducements. Therefore, this research focuses on uncovering the deceptive mechanisms to promote gambling activities across the entire live streaming ecosystem, from gambling streams on platforms like Kick and Twitch to clips circulating on social media. By examining these deceptive mechanisms, this research can help contribute to the intersection of cybercrime and gambling harms research.

Dr Matthew Edwards (Bristol)

Dr Phillip Newall (Bristol)

 Jay Parker

Compiler-induced bugs in perfect software

Some types of code, such as cryptographic algorithms, must always take the same amount of time to execute, independent of the secret material used in the process - otherwise an attacker with an accurate clock could recover the keys or decrypt messages. But modern programming languages can't express this requirement, and so developers use tricks and workarounds instead. This project aims to improve the tooling available for detecting this condition, and implement a few new approaches to help eliminate it, rather than relying on unstable solutions.

Dr Daniel Page (Bristol)

Dr Gretchen Hallett (Bristol)

Michael Tonkin 

A Pathway and Resilience-Centric Dynamic Risk Assessment Model for Operational Technology Systems

My research focuses on dynamic risk assessments (DRA) for operational technology (OT). Specifically, I am developing a new DRA that is more in-line with OT system defender's typical mode of thinking - that being network pathway resilience. Furthermore, there are many DRA solutions for IoT, but significantly less for OT. A part of my research involves looking at these solutions and seeing where they may be used to fill gaps in OT DRA solutions. Finally, I am considering how the rise of agentic AI in offensive cybersecurity might be mitigated to some extent by a newly developed DRA. My research is currently at the early literature review stage, so changes are possible where new information comes to light.

Dr Joe Gardiner (Bristol)