Abstract

Cyber security attacks and threats are prone among the websites used by people now a days. The most vicious threat for people surfing on the websites is phishing. Phishing is recognized as a malicious act of cybercrime to deceive users and steal important data such as Banking Credentials, ATM PINs, cards, and CVV numbers. Our proposed method is based on detecting these websites using a Hybrid Ensembler model. The URL is given as input and the feature extraction is done by thirty-one parameters which mainly include URL length, HTTPS request/response, IP address, sub-domain, etc The modal is trained using various machine learning algorithms like decision tree, K-Nearest, logistic regression, naïve bayes and hybrid ensembler and the output is obtained numerically. The model is trained in such a way that it gives more accuracy and efficiency than the existing method. The Hybrid Ensembler is used to pick a particular algorithm that gives better results for a certain dataset. Python functions are written for each feature extraction and the model is trained in google colab. The instructed modal is transfigured to a pickle file using python pickling for deployment in the flask framework. Flask framework and other UI development tools are used to create web applications where the user can enter the URL to check whether it is a Phishing website or a legitimate website.

Authors

B. Kalaimathi
Sri Ramakrishna Engineering College, India

Keywords

Cybersecurity, Machine Learning, Web Development, Phishing Detection, Ensemble