TrustBase

TrustBase is a Flask-based privacy intelligence app that audits a website using:

• policy/TOS language analysis,
• cookie behavior checks,
• optional AI breach snapshot generation,
• and a final averaged grade.

It is designed for fast, practical trust/compliance reviews from a single URL input.

---

Features

• Policy Risk Audit

  • Detects risky privacy language categories (collection, sharing, rights, retention, vague terms).
  • Highlights risky terms in policy text.
  • Produces a policy grade + risk level.

• Cookie Compliance Audit

  • Collects observed cookies for a site.
  • Compares behavior against policy disclosures.
  • Produces cookie compliance score + grade.

• Optional AI Breach Snapshot

  • Toggle-able checkbox to reduce token usage.
  • Generates structured incident summaries with source links and synopsis.
  • Produces breach grade + risk level.

• Final TrustBase Grade

  • Averages available component grades:
    • Policy,
    • Cookie,
    • and optional Breach.

---

Tech Stack

• Python 3.10+
• Flask
• Playwright (for cookie collection)
• Google GenAI SDK (google-genai)
• BeautifulSoup + requests

---

Quick Start (Windows PowerShell)

1. Create and activate virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1

2. Install dependencies

pip install -r requirements.txt

3. Install Playwright browser runtime

playwright install chromium

4. Set environment variables

Create/edit .env:

GEMINI_API_KEY=your_real_api_key_here

5. Run app

python app.py

Open:

• http://127.0.0.1:5000/ (Home)
• http://127.0.0.1:5000/compare (Main audit workflow)

---

API Key Test Utility

You can test your Gemini key via CLI:

python test_key.py

This script auto-discovers models from your account and falls back across candidates.

---

Routes

• / → Home page
• /about → Product/function glossary and grading documentation
• /compare → Primary TrustBase audit page
• /cookie-audit → Alias of the same audit workflow
• /test-key and /test_key.html → Web key test page

---

Grading Model (Current)

• Letter-to-points mapping: A=4, B=3, C=2, D=1, F=0
• Final grade = average of available component grades
• Final risk level:
  • A/B → Low
  • C → Medium
  • D/F → High

Component grades include:

• Policy grade (from policy risk score)
• Cookie grade (from cookie compliance score)
• Breach grade (if AI breach lookup is enabled)

---

Project Structure

• app.py — Flask app, routes, grading aggregation, AI breach integration
• Parser.py — policy term detection and risk scoring logic
• CookieAudit.py — cookie classification, policy fetch, cookie truthfulness logic
• test_key.py — CLI Gemini key/model connectivity test
• templates/ — Jinja templates (base, index, about, compare, etc.)
• static/style.css — global UI styling
• static/trustbase.png — logo asset
• requirements.txt — Python dependencies

---

Notes / Limitations

• AI breach summaries may vary by model output and available public info.
• Always validate critical legal/compliance findings with human review.
• Some sites aggressively block crawling/automation; cookie or policy fetch can fail per-site.

---

License

No license has been specified yet.