PhishGuard AI

01_system_architecture
02_detection_workflow
03_real_application
04_browser_extension
05_reporting_blacklist
06_legitimate_url_scan
07_suspicious_url_scan
08_high_risk_detection
09_browser_extension_scan
10_pre_navigation_security_warning

Inspiration

Phishing is a common cybersecurity threat in which attackers create deceptive websites to trick users into sharing sensitive information such as usernames, passwords, and financial details. Some phishing websites look similar to genuine websites, making them difficult for users to identify. We were inspired to work on this project to explore how artificial intelligence and machine learning can help identify suspicious websites and improve online safety.

What Our Project Does

PhishGuard AI is a Real-Time AI/ML-Based Phishing Detection and Prevention System. The aim is to analyze website URLs and selected website characteristics to identify potentially malicious websites. The proposed system classifies website risk into three categories: Safe, Suspicious, and Phishing. Based on the assessed risk, it aims to allow safe websites, warn users about suspicious websites, and block websites identified as phishing.

How We Plan to Build It

Our approach includes collecting legitimate and phishing URL data, preparing the dataset, extracting useful URL features, and analyzing selected domain and webpage characteristics. We plan to train and compare machine learning algorithms such as Decision Tree, Random Forest, Logistic Regression, Support Vector Machine, and Naive Bayes to select a suitable model.

The system will use relevant features such as URL length, suspicious keywords, special characters, subdomain structure, and other website indicators. We also plan to develop a browser-oriented interface to display risk information and provide warnings to users. Model performance will be evaluated using accuracy, precision, recall, F1-score, false-positive rate, and processing time.

What We Learned

Through this project, we learned about phishing attacks, malicious URLs, dataset preparation, feature extraction, machine learning classification, and cybersecurity system design. We also understood that HTTPS alone does not guarantee that a website is safe and that a single suspicious feature is not enough to confirm a phishing attack. Multiple relevant indicators should be considered during risk assessment.

Challenges We Identified

One of the main challenges is identifying newly created phishing websites that may not appear in existing blacklists. Another challenge is reducing false positives, where legitimate websites may be incorrectly classified as malicious. We also need to consider detection speed and the computational resources required for real-time analysis.

Future Scope

In the future, the project can be improved by expanding the dataset, evaluating different machine learning models, improving webpage analysis, and developing a more complete browser extension. Additional testing can help assess the system's accuracy, reliability, and response time.

Our goal is to develop a practical and user-friendly phishing detection system that combines machine learning and website analysis to help users browse more safely.

Built With

Created by

I led the development and integration of PhishGuard AI, a real-time AI/ML-based phishing detection and prevention system. My primary responsibilities included designing the overall system architecture, developing the backend API, integrating the machine learning model, and coordinating the implementation of the browser extension.

I worked on URL analysis, feature extraction, risk scoring, and classifying websites into safe, suspicious, and phishing categories. I also contributed to integrating the XGBoost machine learning model with the backend so that URL analysis results could be returned to the extension. In addition, I worked on API testing, URL normalization, debugging classification inconsistencies, and verifying the consistency of risk scores displayed in the extension.

I coordinated system-level testing, including backend regression tests and controlled browser-extension testing, to verify that high-risk results could trigger a warning page. Through this project, I gained practical experience in machine learning integration, cybersecurity, backend development, debugging, and end-to-end system validation.

UTHAYAGIRI E
My role in PhishGuard AI focused on developing the user-facing interface and browser-extension experience. I contributed to designing the interface so users could submit a website URL for analysis and understand its assessed risk through clear classifications, risk scores, and warning messages.

My responsibilities included working with HTML, CSS, and JavaScript to build and refine the interface, connecting frontend components with backend API responses, and helping present detection results in a user-friendly format. I also contributed to the browser extension's popup experience and the presentation of safe, suspicious, and high-risk website results.

I focused on making the application easier to use and understand, ensuring that risk information and warnings were presented clearly. This work helped me develop practical skills in frontend development, API integration, browser-extension development, and building security-focused user interfaces.

SIVASELVAKUMAR M
My role in PhishGuard AI focused on supporting the machine learning and data preparation aspects of phishing detection. I contributed to understanding phishing and legitimate URL datasets, identifying useful URL characteristics, and supporting the preparation of data for machine learning-based classification.

My responsibilities included examining features such as URL length, suspicious keywords, special characters, subdomain structure, and HTTPS usage, and understanding how these characteristics can help distinguish potentially malicious websites from legitimate ones. I also contributed to evaluating the importance of model performance measures such as accuracy, precision, recall, F1-score, and false-positive rate.

I focused on understanding the challenges involved in detecting previously unseen phishing websites while reducing incorrect warnings on legitimate sites. Through this project, I developed practical knowledge of dataset preparation, feature engineering, machine learning evaluation, and cybersecurity threat detection.

VUNGARALA PRANEETH VS

Updates

UTHAYAGIRI E started this project — Oct 09, 2026 05:51 PM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.