| title: PDAT Phishing Detection And Awareness Tool | |
| emoji: 🎣 | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: gradio | |
| sdk_version: "4.44.1" | |
| app_file: app/app.py | |
| pinned: false | |
| # PDAT — Phishing Detection and Awareness Tool | |
| Deployment repository for the PROM02 dissertation's PDAT prototype (Objective 3) and evaluation study (Objective 5). Serves the fine-tuned DistilBERT classifier (Objective 2 winner) with LIME-based explanations behind a Flask web interface, plus the pretest/tool-interaction/posttest/SUS/NASA-TLX study flow. | |
| This Space runs the Flask app directly (not an actual Gradio interface) on the free CPU Basic tier, using the Gradio SDK only because Hugging Face gates the Docker SDK behind a paid plan. `app/app.py` binds to port 7860 and serves all routes itself; no `gradio` import is required or used at runtime. | |
| Routes: | |
| - `/` — standalone classify demo | |
| - `/study` — the evaluation study flow | |
| - `/health` — health check | |
| ## Data persistence | |
| Study responses write to local CSV by default. Since this Space's filesystem is ephemeral, the Google Sheets mirror (`GOOGLE_CREDENTIALS_JSON` and `GOOGLE_SHEET_ID` Space secrets) must be set before real participants use this, or response data can be lost on a restart. | |
| ## Source | |
| Application code develops in the private repository at [Ad163/phishing-detection-awareness-tool](https://github.com/Ad163/phishing-detection-awareness-tool). This Space repository is deploy-only and additionally bundles the trained model weights, which are excluded from the code repository. | |