Scribd
Upload
17 views13 pages

Ddu Project

The document outlines an AI-powered phishing website detection system that utilizes machine learning, heuristics, and APIs for real-time detection. It aims to provide users with proactive protection against phishing attacks through a browser extension and a transparent, explainable AI model. The project emphasizes the need for innovative solutions to combat increasing phishing threats and includes plans for future enhancements such as deep learning integration and user-friendly alerts.

Uploaded by

chandanvishwa15
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
17 views13 pages

Ddu Project

The document outlines an AI-powered phishing website detection system that utilizes machine learning, heuristics, and APIs for real-time detection. It aims to provide users with proactive protection against phishing attacks through a browser extension and a transparent, explainable AI model. The project emphasizes the need for innovative solutions to combat increasing phishing threats and includes plans for future enhancements such as deep learning integration and user-friendly alerts.

Uploaded by

chandanvishwa15
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

“AI Powered

Phishing Website
Detection
System”
Real-time detection using Machine Learning,
Heuristics & APIs
Meet Our Team
“Collaboration is the key to
innovation.”

Ayush
Aniket Jaiswal Vishwakarma Anupam Tiwari
Frontend Developer & Project Lead & Machine Backend Developer &
UI/UX Designer Learning Developer Integration Specialist
Problem Statements
 Traditional detection
 Phishing websites often depends on blacklists,
look visually identical which fail to catch new
to genuine ones sites

 Users need a real-time,


 Successful attacks lead intelligent system that
to financial loss, data can flag phishing
theft, and diminished attempts at the moment
trust of access
Objectives
 Build a machine learning model to detect phishing
websites
 Use heuristic features and 3rd-party APIs (like Virus
Total & WHOIS) for enhanced detection
 Integrate a real-time browser extension for proactive
protection
 Build a transparent system with explainable AI
 Deploy a full-stack solution with a dashboard and
reporting features
Why This Project Is Unique

 Hybrid Approach  Real-Time Browser  Explainable


Extension Predictions
ML + heuristics + APIs
= high accuracy Warns users before Transparent decisions
they click with LIME/SHAP

 Admin Dashboard  Crowdsourced


Reporting
Log analysis, pattern Users can submit
detection, and visual suspicious links,
stats building a live dataset
System Architecture
This diagram explains the overall flow of how a URL is processed
through various stages of our phishing detection system.
Technology Stack
Frontend Backend
•Chrome Extension
Flask REST API
(JavaScript, HTML, CSS) Firebase for logging
•Streamlit Dashboard
and real-time sync
(Python)

Machine Learning External APIs


•Random Forest, XGBoost,
•WHOIS API (Domain info)
Support Vector Machines
•Feature Engineering with •VirusTotal API (Threat score)
Scikit-learn •Google Safe Browsing API
•Explainability via LIME & (Blacklist check)
SHAP
Dataset & Feature Engineering
Feature Preprocessing
Sources Used Categories: Steps

• URL-Based: Length, use


• PhishTank – of • Clean URLs
Verified phishing ‘@’, ‘//’, subdomains,
• Encode categorical
URLs redirection count
• Kaggle Phishing • Domain-Based: Age of data
domain, expiration time, • Normalize numerical
Dataset – Balanced WHOIS info
• Security-Based: HTTPS
features
labeled data • Handle class
• Alexa Top 1M – present?, certificate valid?
• Content-Based imbalance with
Legitimate websites (optional): Keyword • Smote/undersampling
analysis, HTML tag patterns
Real-Time Browser Extension
This is where your project goes from
“academic” to “real-world.”

Features:
 Auto-scans every URL opened
in Chrome
 Sends the URL to backend API
 Backend returns prediction
(phishing or safe)
 If phishing → blocks the page
and shows red warning screen
 Simple, clean UI integrated
with Chrome browser
Future Scope

 Deep Learning  Voice-Based  Integration into


Integration Phishing Alerts Antivirus or
Incorporating deep learning Help users with disabilities Browsers
into applications enhances via voice notifications Deploy as part of open-
decision making and source security tools
accuracy.

 Telegram/  Continuous
WhatsApp Bot Model Learning
Users can send URLs to a Implementing continuous
bot and get instant verdicts model training ensures the
system adapts and evolves
with new data.
Project timeline

Research + Feature Model Tuning + Chrome


Dataset Engineering + ML Evaluation Extension Dev
Collection Models

Dashboard + Testing, Demo,


Explainable AI Final Report
Conclusion
 Phishing attacks are increasing—and traditional solutions
can’t keep up

 Our system detects phishing proactively using ML,


heuristics, and real-time tools

 With a working browser extension and explainable model,


it’s ready for real-world use

 This project combines AI, Cybersecurity, and Real-Time


Systems — and has the potential to scale as a full product

“Our goal is to not just detect phishing, but to educate


and protect users in real-time.”
Thank You!

“Our goal is to not just detect phishing, but to educate


and protect users in real-time.”

You might also like