Phishing URL Detection System

Overview

A cybersecurity and machine learning web app that classifies URLs as legitimate or phishing using hand-crafted lexical features and a trained scikit-learn ensemble, with explainable per-feature scores.

Problem or Opportunity

Phishing links remain a common attack vector, and teams need practical triage tools that explain why a URL may be suspicious.

Proposed Solution

The system combines a Flask REST API with a React dashboard, providing classification results and feature-level explanations useful for security education and triage workflows.

Key Capabilities

• URL lexical feature extraction
• Ensemble model classification
• Explainable per-feature contribution scores
• React dashboard for interactive analysis
• Reproducible model training artefacts

Architecture and Workflow

Flask serves the REST API and model inference layer. React provides the frontend dashboard. Trained model artefacts are loaded via Joblib for consistent predictions.

Security Considerations

• Input sanitisation for submitted URLs
• No automatic fetching of untrusted links during analysis
• Clear presentation of model limitations

Challenges and Decisions

• Balancing model accuracy with interpretability
• Designing features that generalise across phishing patterns
• Communicating confidence levels responsibly in the UI