Intelligent Phishing Website Detection Using Deep Learning

Artificial Intelligence and Machine Learning technologies have become a crucial tool for the identification and detection of phishing URLs from samples with a reduced inspection time while avoiding human errors due to inconsistency and fatigue. The “blacklisting” approach, which involves the recording of IP address of servers as well as blacklisted URLs to an antivirus database, is a standard and well-known method for detecting phishing sites. To bypass the blacklisting defense layer, attackers employ a variety of sophisticated techniques to deceive users, including altering the URLs to make them look genuine through obfuscation and other methods. However, using such techniques have its own drawbacks such as false-positive results. To beat the downsides of the blacklisting method, security researchers are looking for ways to adapt and work on several Deep Learning technologies. Therefore, it is very important to develop a model that intelligently detects phishing URLs using small training data with higher accuracy. In addition, it is also crucial to increase our efficiency of detection to ensure that our system is free from phishing and malicious URLs. In this paper, we have proposed a neural network named Autoencoder which employs outlier analysis to discriminate and classify websites as genuine or phishing websites. The algorithm analyzes several fake and genuine URLs and study their features to precisely identify the phishing sites including those that are made in real-time also known as zero-hour phishing websites utilizing this approach.