APuML: An Efficient Approach to Detect Mobile Phishing Webpages using Machine Learning.

Area of Science:

• Cybersecurity
• Machine Learning
• Mobile Computing

Background:

• Mobile phone usage and web access have surged, increasing vulnerability to sophisticated cyber threats like phishing.
• Existing mobile security mechanisms and browser capabilities lag behind desktop solutions, necessitating specialized detection methods.
• Current anti-phishing techniques for mobile devices are insufficient, highlighting the need for advanced, comprehensive solutions.

Purpose of the Study:

• To develop and present an efficient approach for detecting malicious mobile webpages.
• To address the limitations of current mobile anti-phishing strategies.
• To provide users with an interactive mobile application for enhanced security.

Main Methods:

• The proposed Anti Phishing using Machine Learning (APuML) approach extracts static and site popularity features from URLs.
• A feature vector is created and analyzed using machine learning classification algorithms.
• The Random Forest classifier was selected for its superior performance in identifying malicious sites.

Main Results:

• The APuML approach, utilizing the Random Forest classifier, achieved a high detection accuracy of 93.85%.
• The system demonstrated effectiveness in identifying various advanced threats, including drive-by downloads, zero-day attacks, and clickjacking.
• An endpoint application was developed for seamless user interaction on mobile devices.

Conclusions:

• APuML offers an efficient and accurate method for detecting malicious mobile webpages, significantly improving mobile security.
• The approach successfully identifies a range of sophisticated attacks, providing a more comprehensive defense than existing methods.
• The developed mobile application enhances user engagement and accessibility for real-time threat detection.