Predicting Phishing URL Using Filter based Univariate Feature Selection Technique

Phishing URL acts as a trustworthy website and cheats the user by taking all the essential data or the personal information. Hence detecting phishing attack is a very tedious task.Though the organizations are working hard to mitigate this threat still it is the prime concern to alert end users awareness in preventing such threat because phishers often exploit users by visual similarity of the phishing website over the legitimate website.Therefore, our objective is to design a detection framework using feature selection methods of machine learning technique. Univariate feature selection methodology is one of the best pre-processing technique for choosing the most relevant and important feature because it adopts the statistical test which yields the statistical significance between the features for determining the predictive power. It also enhance the learning power of the model in less time. In our experiment, we have used the three univariate feature ranking algorithms such as Information Gain (IG), Chi-square and ANOVA to select the best features that evaluates the supervised machine learning algorithms kNN (k-nearest neighbour), DT (Decision Tree), LR (Logistic Regression) and NB (Naive Bayse). Our experiment demonstrates that, kNN outperforms all and the accuracy is of 99.8%.