Feature Selection & ML Based Prediction of Phishing Websites

EasyChair Preprint 8412

Abstract

The Internet and cloud based technologies increase online purchases and transactions, in recent years. Phishing be a well-known assault that deceives consumers into seeing harmful material in exchange on the behalf of their personal information. However, due to ineffective security systems, the number of victims will grow exponentially. The Internet's anonymous and unregulated foundation makes it more vulnerable to phishing attempts. In this paper, we use a point wise mutual information method to offer a phishing detection system that uses features from the website URLs. We built a supervised machine learning system for phishing website detection. We used “Embedding, Sentiment, and Lexicon characteristics, as well as PMI-semantic orientation”, in the study. The methods “SVM, Naïve Bayes, KNN, Decision Tree and algorithm Random Forest,” were used to apply extracted features. Experiments using our suggested framework in a multi-class scenario, as well as in a binary setting, show promise in terms of “values of Kappa, enhanced accuracy, and calculated f-values”. These findings suggest that the framework we've provided be a viable option on the behalf of detecting malicious phishing behavior with severity of online links of social networks and other fake web URLs. Finally, we used different machine learning techniques to compare the outcomes of suggested and baseline characteristics. 10 fold cross validation calculated 90.363 highest accuracy and all four experiment evaluated always high accuracy on the behalf of Random Forest on the behalf of training dataset on 80%. The test result also calculated enhanced accuracy on the behalf of Random Forest on 20% or test dataset.

Keyphrases: Algorithm Naïve Bayes, Decision Tree (J48), KNN, Random Forest algorithm and, SVM

@booklet{EasyChair:8412,
  author    = {Anjaneya Awasthi and Noopur Goel},
  title     = {Feature Selection & ML Based Prediction of Phishing Websites},
  howpublished = {EasyChair Preprint 8412},
  year      = {EasyChair, 2022}}