We are an assurance company, so our own posture has to survive the scrutiny we apply to others. We are working towards Cyber Essentials; we do not yet hold it and claim no certification.
Email inquiries@ukaievaluation.com with the subject SECURITY, including
reproduction steps and impact. We acknowledge within five working days. Please do not
open a public issue for a security report.
- It reads CSV files of scores and labels and writes a JSON file. It makes no network calls, executes no model, and imports nothing outside the Python standard library.
- It never needs model weights, images or patient-identifiable data. If you are
evaluating a clinical model, keep the optional
groupcolumn pre-coded (opaque stratum labels) unless a data processing agreement covers it. - Treat every prediction file as sensitive regardless. Keep them out of this
repository:
.gitignoreexcludes the common data extensions, and CI runs a secret scan on every push and pull request.