NTPhish: A CNN-RNN Hybrid Deep Learning Model to Detect Phishing Websites

Abstract

It is certainly peculiar that even after securing one's infrastructure with state of the art technologies, companies still get compromised. The question thus arises ‘How is an attacker able to circumvent such sophisticated defenses?’. The answer to this is relatively simple. Attackers exploit the most vulnerable component in the chain, also known as humans. They do so by targeting people with fake emails and websites. Attackers often spoof legitimate services and modify them to perform nefarious activities. This act of portraying a malicious resource as a legitimate resource is known as phishing. The main motive behind phishing is to trick the victim into revealing personal information or more often phishing acts as a precursor to malware infections. Advancement in technology has made it easier for attackers to spoof a legitimate resource with almost zero flaws. It makes it extremely difficult for the victims to evade such attacks. However with the aid of artificial intelligence detecting such websites becomes extremely easy and accurate. In this research we propose a hybrid deep learning model to detect phishing websites. The hybrid model is a combination of CNN and RNN algorithms and gives a high degree of accuracy in phishing website detection. For training and validation the datasets have been used. The results of our experiments show that the proposed model performs better than traditional deep learning models.