An Intelligent Security Intervention system designed to protect mobile banking users from phishing attacks using XGBoost Machine Learning. This system analyzes URL structures in real-time to identify malicious links before users provide sensitive credentials.
- High Accuracy: 97.06% detection rate on the UCI Phishing Dataset.
- Low Latency: Average inference time of 15.58 ms, optimized for mobile user experiences.
- Algorithm: Gradient Boosted Decision Trees (XGBoost).
- Deployment: Live web interface built with Streamlit.
The system extracts 30 features from a submitted URL, categorizing them into:
- Address Bar Features: IP addresses, URL length, and shortening services.
- Abnormal Features: Redirecting symbols (
//) and prefix/suffix hyphens. - Security Markers: HTTPS/SSL presence and subdirectory depth.
- Language: Python 3.12
- ML Library: XGBoost, Scikit-Learn
- Interface: Streamlit
- Environment: Anaconda / Virtualenv
- Clone the repository:
git clone https://github.com/CodeeSam/mobile-phishing-detector.git