A Prediction of Phishing Websites by Optimal Feature Extraction using Recurrent Neural Network
Niroshini Infantia H, Saira Banu Mohammed Rasool, M Gnanaprakash, M. Senthilmurugan · 2022 6th International Conference on Trends in Electronics and Informatics (ICOEI) · 2022
Phishing is a cyber-attack threat that tries to access private information like credit card information, personal information and passwords. In the current scenario most of the online intrusions involves phishing. For digital surveillance assaults the number crosses above 78%. Yet, these sites can be related to their substance and internet browser based data can be utilized for order. The features are extracted in a comprehensive manner from different perspectives, but still there is a drawback is that training with all these extracted features tends to over-fit the model to the given training dataset. This overfitting is an effect of the small influence features that traps the neural network model. Apart from this, the training process also depends on 3rd party applications like WHOIS lookup and Page rank database. To eradicate these limitations we have proposed a website for detecting phishing sites where this approach is based on an optimal feature selection algorithm to select the essential features. The feature value index (FVI) is used to evaluate the impact and the sensitivity of the feature on the model. Then the proposed algorithm is used to select only the optimal features. The selected optimal features are then fed to the neural network to train the learning model which is the optimal phishing website prediction model. The model then predicts the probability of a website being a phishing website when its corresponding URL is fed in to the model.