Phishing Detection System Using Dueling Network

Ikram, Mohammed Afnan (2020) Phishing Detection System Using Dueling Network. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (836kB) | Preview
Preview	PDF (Configuration manual) Download (508kB) | Preview

Abstract

Cyber attackers pull millions of dollars through phishing attack every year. Adding basics of social engineering in phishing makes the attack more effective. We can say, it is the modest way of cybercrime which has the aim of playing foul with people to steal their private sensitive information like bank details, passwords, credit/debit card details or other important personal identification details. The personal data or credentials stolen are used by the attacker to get illegal access of victim’s personal accounts which could result in leak of private data or monetary loss. so, the first step taken to initiate the phishing attack is to send the infected messages and gather victims information, than on the basis of the information gathered through social engineering, attacker setup the deceptive copy of the original website, where the target is conned to enter its personal information or credentials. In today’s era of Artificial intelligence, machines are getting more advanced, Intellectual and smart enough to take decision on their own. It will not be wrong assuming that using these advancement, cyber criminals are also working hard finding loopholes in our system which can be undetected. So, it is very important to develop a technology which is smart enough to evolve itself by leaning different pattern and function for detecting phishing websites. Many researches are done and various phishing detecting systems and tools are developed using different machine learning algorithms. taking inspiration from the vigorous approach and evolving nature of these phishing pages, in this paper, a novel approach is introduced using Random forest and dueling network of reinforcement learning model, where machine learns from training dataset and show improvement in accuracy results. The model in this research shows the capabilities of learning and adapting the unstable dynamic nature of the phishing websites and hence adopt the detected features related to the phishing website. The model also improves itself by learning from different data input. Dueling network is used in this research, where 2 Q-learning models are used to increase its efficiency and accuracy. The model also works on rewards-based system, where the model is awarded with the reward of 10 credits, if it performs well in detection, which thrives it to improve more. The proposed work showed high accuracy rate of 96%, and has further scope of increasing its accuracy with increase in model training.
Keywords: Reinforcement, heuristic, phisher, Random forest, dueling

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
Divisions:	School of Computing > Master of Science in Cyber Security
Depositing User:	Dan English
Date Deposited:	26 Jan 2021 16:01
Last Modified:	26 Jan 2021 16:01
URI:	https://norma.ncirl.ie/id/eprint/4495

Actions (login required)

View Item