Machine Learning Algorithms for Detecting Phishing Websites

K P SENTHILKUMAR M.E.,C.S.E.,
Assistant Professor, Dept of AI&DS,
Kings Engineering College, Chennai-600016.

ABSTRACT
The simplest method of obtaining sensitive information from unwitting people is through a phishing attack. The goal of phishers is to obtain crucial data, such as username, password, and bank account information. People working in cyber security are currently looking for reliable and consistent methods of detecting phishing websites. In order to distinguish between legal and phishing URLs, this article uses machine learning technology. It extracts and analyzes many aspects of both types of URLs. Algorithms such as Support Vector Machine, Decision Tree, and Random Forest are used to identify phishing websites. By evaluating each algorithm’s accuracy rate, false positive and false negative rates, the study aims to identify phishing URLs and identify the best machine learning method.
Keywords
Phishingattack,Machinelearning
1. INTRODUCTION
Due to how simple it is to develop a phony website that closely resembles a legitimate website, phishing is now a top worry for security researchers. Although experts can spot fraudulent websites, not all users can, and as a result, some users fall prey to phishing scams. The attacker’s primary goal is to obtain login information for bank accounts.Businesses in the US lose $2 billion annually as a result of their customers falling for phishing schemes [1]. The annual global cost of phishing was pegged at $5 billion in the third Microsoft Computing Safer Index Report, which was published in February 2014 [2].

Due to a lack of user awareness, phishing assaults are becoming more successful. Since phishing attacks take advantage of user vulnerabilities, it is highly challenging to mitigate them, but it is crucial to improve phishing detection methods. The “block list” method, which is the standard technique for detecting phishing websites, involves adding rejected URLs and Internet Protocol (IP) addresses to the antivirus database. Attackers modify the URL to appear authentic by obfuscation and many other straightforward ways, such as fast-flux, in which proxies are automatically constructed to host the website, algorithmic production of new URLs, etc., to dodge blacklists.

A zero-hour phishing assault can be detected using heuristic-based detection, which contains features that have been shown to exist in phishing attacks in reality, but the existence of these traits is not always guaranteed in such attacks, and the false positive rate for detection is very high [3].Many security experts are now focusing on machine learning techniques to overcome the limitations of blacklist and heuristics-based methods. Machine learning technology is made up of numerous algorithms that use historical data to forecast or make decisions about future data. This method uses an algorithm to examine a variety of genuine and blacklisted URLs and their characteristics in order to precisely identify phishing websites, including zero-hour phishing websites.




machine-learning-algorithms-for-detecting-phishing-websites
FREE DOWNLOAD

@ engpaper.com published paper
PUBLICATION PROCEDURE WITH US ENGPAPER.COM
ENGPAPER.COM PUBLISHED PAPERS