Code ▾Python 3 (ipykernel)

QBlock

Built at Western Cyber Society

A defence against quishing: phishing through QR codes. Scan a code with QBlock and it classifies the destination site as safe or malicious before you visit it. I co-developed it with two teammates, and we demoed it at Canadian Tech Summit 2026.

Python
XGBoost
BERT

LinkedIn post GitHub

Nathan and his two QBlock teammates standing beside their QBlock project poster, with a laptop running the demo on the table

In [1]:

qblock.summary()

Out[1]:

team: Western Cyber Society
members: Cameron Kroupa, Christian Uhm, Nathan Feng
role: Machine Learning Developer
model: XGBoost + BERT ensemble
demoed: Canadian Tech Summit 2026

How it works

A hybrid detection pipeline has XGBoost score structural page features such as tag counts and redirects while a BERT-based model reads the page's text, and an ensemble combines the two into a verdict.

QBlock's Pipeline page following a QR redirect to a fake login page, then listing the page's HTML features and extracted text — Example breakdown: the URL, HTML features and page text

The same breakdown continued: Voter A at 94% with its top feature contributions, Voter B at 81%, and a final verdict of phishing at 93% — Example breakdown: both voters and the final verdict

In [2]:

qblock.evaluate()

Out[2]:

accuracy: 87%
sites_tested: 10,000+
model: XGBoost + BERT ensemble

QBlock

How it works

Keyboard Shortcuts

Notebook

Workspace