✦Web, AI, Python, Machine Learning · 2026
PhishGuard
A hybrid phishing detection system that pairs a rule-based expert engine with XGBoost and Random Forest — and explains every verdict instead of just issuing one. 95.71% accuracy across 11,430 test samples.
Overview
1. A phishing link that actually works looks completely ordinary. That is the whole problem — and it is why a black-box classifier is a weak answer. A system that tells a user "this is dangerous" without saying why teaches them nothing, and eventually they stop listening.
2. PhishGuard is deliberately hybrid: a rule-based expert system using forward chaining, running alongside XGBoost and Random Forest models. Ninety-one engineered features drawn from the URL and the page content feed both sides. Every verdict carries the reasoning behind it, built on explainable-AI principles.
3. On the Web Page Phishing Detection Dataset it reached 95.71 percent accuracy across 11,430 test samples.
Highlights
- 0195.71% accuracy on 11,430 test samples
- 0291 engineered features from URL and page content
- 03Forward-chaining expert system fused with ML classifiers
- 04Explainable AI — every classification is traceable
- 05XGBoost and Random Forest evaluated head to head
Screens
Click to enlarge
Next project
Crew Management IS PT PELNI (Persero)
Enterprise Web App · 2025