02 / CASE STUDY

Wutong Defense Console

An AI-powered campus telecom fraud detection system combining risk portraits, rules, and machine learning.

Wutong Defense Console visual

Overview

A privacy-aware defense console for identifying telecom fraud patterns targeting students in Hong Kong.

Problem

Fraud signals are distributed across identity, exposure, behavior, and evolving attack patterns.

Solution

Combine a risk-triangle scorer, seven-rule engine, XGBoost, Isolation Forest, and human review.

Technology

Python, XGBoost, Isolation Forest, Streamlit, Differential Privacy

  • Python
  • XGBoost
  • Isolation Forest
  • Streamlit

Architecture

Feature engineering feeds risk portraits, rule-based classification, and hybrid ML detection modules.

Outcome

A system that reports 0.89 AUC in the project benchmark and supports explainable review flows.