Fractional-ewa based deep cnn for phishing attack detection

Abstract

Phishing attack is acclaimed as one of the recognized cybercrime attacks over the internet and mail users. Phishing is a form of unauthorized access of confidential information, like passwords, user names and credit card details. Detection of phishing attacks and classifying the mails still remains a challenging issue. This research presents an effective strategy by developing a newly proposed method called Fractional-EarthWorm Algorithm (EWA) based Deep Convolutional Neural Network. The Fractional-EWA is derived by inclusion of fractional calculus concept to EarthWorm Optimization. The features are extracted using the term frequency and the feature is selected using the Levenshtein distance. The DCNN is trained by exploiting the proposed Fractional-EWA. However, this algorithm achieved maximum accuracy, maximum sensitivity, and maximum specificity of 0.781, 0.782, and 0.722 respectively for chunk percentage of data and achieved the maximum accuracy, maximum sensitivity, and maximum specificity of 0.744, 0.725, and 0.723, respectively for number of features.