The University of Kent, an NCSC (National Cyber Security Centre)-recognised Academic Centre of Excellence in Cyber Security Education (ACE-CSE) and in Cyber Security Research (ACE-CSR), hosted a cyber security research intern – a holder of a NCSC CyberFirst bursary – at the Institute of Cyber Security for Society (iCSS) and School of Computing for 9 weeks over the summer of 2026.
The intern who completed the internship had just completed their first year of studies at Royal Holloway, University of London. Under co-supervision of Dr Virginia Franqueira and Professor Shujun Li, they explored capabilities of multi-modal large language models (MMLLMs) (i.e., Llama 4 Scout and Llama 4 Maverick) to detect images generated by StyleGAN, experimenting with different parameters such as temperature values (that determine the level of determinism involved in the results). They then experimented with Gemini 2.5 and GPT 4.0 to evaluate their capability to generate images from text prompts and differentiate real from fake and the quality of explanations given by the MMLLM to justify its decision. This work is currently being developed further as part of an UG final-year project at the University of Kent.
Dr Virginia Franqueira said: “It was a real pleasure to supervise this talented student over the summer. They remained highly engaged throughout and did not allow the substantial learning curve to become a de-motivating factor. The intern had never even used an LLM before and had to face the challenges of automating a pipeline to perform experiments at scale. Well done; I wish the intern plenty of success in the future!”
Professor Shujun Li said: “The bursary holder’s work formed a natural part of our increasing research, education, and wider impact activities at iCSS on mis-/disinformation and deepfakes. In addition to producing some preliminary evidence on MMLLM-based detection of deepfake images, the work undertaken also led to some new insights on human detection of deepfakes, which was pleasantly unexpected!”
The CyberFirst bursary holder said: “The time I spent at the University of Kent over the summer undertaking my cyber security internship was a very valuable and enjoyable experience, especially thanks to the support and guidance of my supervisors Dr Virginia Franqueira and Professor Shujun Li. I learnt so much about LLMs in such a short time, and successfully automated a testing pipeline using Python scripts which I am very proud of. It was very rewarding when, at the end of my placement, I could see trends in my results between how well different LLMs performed at determining genuine and deepfake images. This opportunity has broadened the horizons of what I consider to be a potential career path—I could definitely see myself doing more research in the future!”