Research
Technology against gender-based violence.
My research interest is technology-facilitated gender-based violence (TFGBV): how technology enables abuse, and how it can help stop it. I focus on two threads. First, AI-generated non-consensual intimate imagery (AIG-NCII), where I work on image provenance and detection. Second, AI used in family-violence contexts, where getting the assessment wrong can leave a survivor without protection.
Building
What I'm working on
- What is this image? An image-authenticity inspector. It reads C2PA provenance to show who made a picture, with what tool, and whether AI was involved. Watermark and metadata checks are next. In progress
- AI-content detection Passive ML detection of AI-generated images: an evaluation framework first, then a model, to find out where detection holds up. Early days
- LLM robustness eval A Best-of-N jailbreak-robustness evaluation for language models, measuring how their safety behaviour holds up under repeated adversarial sampling. In progress
Reading
What shaped this
Li Qiwei, Wells Lucas Santo, Sarita Schoenebeck & Eric Gilbert, “Position: AI/ML Deepfake Research is Misaligned with AI Generated Non-Consensual Intimate Imagery (AIG-NCII)” (ICML 2026).
It's why I treat AIG-NCII as a separate problem from deepfake detection.
Other tinkering
Other things I've built
- Media server A self-hosted Jellyfin media server I run on a VPS, with the Radarr, Sonarr and Prowlarr stack behind Traefik. Ansible bootstraps the server and all services run in their own Docker container. Self-hosted · Ansible
- Watchman A proof-of-concept for my Engineering thesis that secures drone-to-controller comms over a second encrypted channel, with continuous authentication and network intrusion detection. Security · Python
- Lymphedema compression paper Co-authored research on pressure profiling for next-generation lymphedema compression treatment. Paper · 2018