phission [completed]
From HCI prototype to a reproducible demo
A synthetic-only email safety demo for pausing, inspecting destinations, and understanding uncertain results.
- Python
- Reflex
- Pytest
- Playwright
- Synthetic inbox
- MIME + link parsing
- Simulated result + guidance
Problem
Unexpected messages can pressure people into following links or sharing secrets. Phission began as a 2023 HCI/affective-computing prototype; the refresh makes its email-inspection workflow reproducible without connecting a real inbox.
Approach
Bounded MIME parsing extracts readable text and HTTP(S) destinations without rendering email HTML. Typed validation distinguishes valid scores, including zero, from unknown and error states. Per-session state binds results to the selected message and link; keyboard-friendly controls and text explanations support inspection.
Engineering trade-off
Authored emails and simulated reputation results avoid mailbox and provider dependencies but do not establish phishing-detection accuracy. Parsing is deliberately limited; read-aloud is not included. Original usability findings are unavailable; no historical metrics or measured benefits are claimed.
Evidence
115 Python tests on 3.13/3.14 cover parsing boundaries, score validation, and session isolation. Seventy browser checks include four automated accessibility scans. These validate the demo, not production mail security, full accessibility conformance, or physical-device support.