Web Spoofing Prevention: Machine Learning Based Client-Side Defence

• Published in: 2024 2nd International Conference on Sustainable Computing and Smart Systems (ICSCSS) pp. 1098 - 1104
• Main Authors: Chanakya, G., Spoorthi, Mareddy, Kumar, Racha Sanjay, Reddy, Jakka Yashwanth, Chowdary, B.V., Kumar, Reddy Dheeraj
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.07.2024
• Subjects: Fraudulent login pages; Machine learning algorithms; Phish Catcher; Phishing; Random Forest classifier; SQL injection; Surveys; Systematics; Uniform resource locators; web features analysis; Web pages
• DOI: 10.1109/ICSCSS60660.2024.10624881

In today's digital landscape, safeguarding users' sensitive information like passwords and PIN codes presents a formidable challenge in cybersecurity. Every day, billions of users encounter deceptive tactics aimed at luring them to fraudulent login pages, where their confidential data is solicited. These tactics range from phishing emails and enticing advertisements to more sophisticated attacks like click-jacking, malware, SQL injection, and session hijacking. One particularly insidious form of cyber threat is web spoofing or phishing, where attackers create counterfeit versions of legitimate websites to trick users into divulging their private information. To address these challenges, this study has developed a client-side defence mechanism leveraging advanced machine learning techniques. The proposed solution aims to detect spoofed web pages and shield users from phishing attacks effectively. As a proof of concept, this study has introduced a Google Chrome extension called PhishCatcher, which integrates the machine learning algorithm. This algorithm analyses a range of web features and employs a Random Forest (RF) classifier to determine whether a login page is likely to be spoofed. To validate the efficacy of our extension, we conducted multiple experiments using real web applications. The results speak volumes: an impressive accuracy and precision rate of 98.5% were achieved from trials involving 400 classified phishing URLs and 400 legitimate URLs. Moreover, to assess the tool's responsiveness, this study has conducted latency tests on forty phishing URLs, revealing an average response time of just 62.5 milliseconds. Overall, the innovative approach not only offers a robust defence against phishing attacks but also demonstrates remarkable efficiency in safeguarding users' online credentials.