Aarshdeeep Singh J Nandra

Large language models have made typed homework an unreliable indicator of authentic student reasoning. This research proposes VoxProof, a browser-based system that shifts written responses from typing to speech. Spoken responses are transcribed, cryptographically signed, and embedded with invisible verification markers that can later be detected and validated by the browser on the page. By combining voice input, backend verification, and browser-side detection, this project explores a practical method for preserving authorship authenticity in educational settings where AI-generated text has weakened trust in conventional written submissions.