Imbroglio
Overview
Imbroglio reads a website's privacy policy and terms of service, extracts specific data practices with verbatim supporting quotes, scores them against a published, versioned rubric, and monitors them for changes. It's a four-person team project for CS 4360 Senior Experience at MSU Denver, where I'm both technical lead and project lead. The prototype finds and stores every version of a site's policies, analyzes new sites through a job queue and a separate worker service on Railway, and serves results through a versioned read API. A public website and a browser extension come next.
Why It Exists
Nobody reads privacy policies, and the companies writing them know it. Imbroglio does the reading and shows what a site actually does with your data, with the quotes to back it up, so people can make an informed choice without a law degree.
Features
- ·Policy discovery: link text, URL patterns, common paths, iframe follow, and a headless browser fallback
- ·Every document version extracted, hashed, and stored so changes can be tracked
- ·Job queue with a dedicated worker service on Railway; the web service never fetches a site itself
- ·Versioned read API with search, per-client rate limits, and a daily cap
- ·Privacy-preserving lookup by hash prefix, so clients never reveal which site a user is on
- ·Audit page where reviewers verify results, producing a measured accuracy figure
- ·Outbound fetch guard that blocks private, loopback, and cloud metadata addresses
- ·Next up: LLM extraction with quote verification, rubric scoring, a website, and a browser extension