Working Papers

Working papers and methodology notes from the MLNavigator Research Group. These are early lab materials and may change as the work matures.

security data compliance draft

2026-01-08

Proprietary Data Exposure in AI Systems

Working draft outlining how proprietary data can leak through AI systems, with mitigation ideas for enterprise use.

Read note →

verification offline cryptography draft

2025-11-12

Offline Inference Verification: A Hash-Chain Approach

Draft note on a hash-chain approach for verifying inference runs in air-gapped environments. We describe a proposed artifact format and signing protocol, with early prototype measurements in progress.

Read note →

energy measurement apple-silicon draft

2025-09-20

Joules Per Token: An Energy Measurement Methodology for Apple Silicon

Draft methodology note outlining how we measure inference energy on Apple silicon using macOS powermetrics. The protocol is intended for repeatable internal benchmarking; tooling and results are still in development.

Read note →