Phishing Techniques Detection System

Abstract

Phishing has become the most common and harmful cyber attacks of our current digital era. Phishers exploit users' confidence in online services by creating fake websites that closely resemble authentic sites such as banks, online shopping websites, and government websites. Such fake sites are created to trick users into sharing sensitive information including usernames, passwords, credit card numbers, and other identifying information. Traditional detection methods, such as checking URLs against blacklists or scanning for suspicious keywords, have proven ineffective. They are often unable to identify new phishing pages or imperceptibly constructed attacks that adapt to evade existing defenses.
Here, we present a more comprehensive approach that takes into account beyond mere surface features like URLs or domain names. Our phishing-detecting mechanism is focused on the inspection of the underlying form and behavior of the sites themselves, drawing features from HTML elements and user interactions to augment the understanding of potential threats. Through machine learning techniques—specifically Random Forest algorithms—we are able to classify websites more accurately based on suspicious trends in forms, scripts, iframes, and redirects.
We harvested and processed thousands of real and phishing websites to create a robust dataset that replicates real-world scenarios. Our system was also trained to detect minor web design and behavior changes that are typically detected by conventional techniques. The results indicate significant improvements in accuracy, recall, and precision over existing techniques.
This work is one part of ongoing endeavor to make the internet a safer place through having a proactive, adaptable, phishing detection system. It demonstrates the capability of applying the incorporation of structural and behavioral analysis and machine learning models towards providing a stronger defense for counteracting perpetually evolving cyber threats. The system is resource-light, scalable, and platform-agile, hence allowing users to surf the internet with greater assurance.