Scam Buzzer v1.0.0
Initial release of Scam Buzzer, a Chrome extension that detects suspicious URLs using a local machine-learning model.
Features
- Analyzes URLs for phishing, malware, defacement, and benign classifications
- Displays LOW, MEDIUM, or HIGH risk levels
- Shows warnings directly on suspicious pages
- Runs entirely on your computer through a local FastAPI server
- Does not send URLs or browsing data to external services
- Includes a configurable phishing-risk threshold
- Trusted domains can bypass model predictions
Performance
Evaluation on domain-grouped test data:
- Accuracy:
0.897 - Macro F1:
0.812
These results are more realistic than random URL splits, which can overestimate performance because URLs from the same domain may appear in both training and test sets.
Installation
-
Install Python 3.10 or newer.
-
Install dependencies:
pip install -r requirements.txt -
Start the local API:
python server_entry.py
-
Open
chrome://extensions. -
Enable Developer mode.
-
Select Load unpacked and choose the
extension/folder. -
Open a website and click the Scam Buzzer extension icon.
The extension connects to http://localhost:8000 by default.
Important Limitations
This is a learning project, not a production security product.
- Detection is based only on the URL string.
- It cannot inspect page content, redirects, or login forms.
- It may miss new phishing domains or flag legitimate sites.
- The API has no authentication and should only be run on localhost