Effective attribute selection and classification technique for phishing attacks detection

Wireless communications are fast growing technologies. Many online customers have suffered from phishing attacks while accessing the internet services. If customers access a fraudulent webpage using their user ID and secret code, then their credentials have been threatened by the invaders which may be utilized for suspected activities. Generally, a phishing webpage looks very identical to the respective trusted webpage for attracting the vast number of customers. To identify this kind of attacks, different machine learning-based techniques including Support Vector Machine (SVM), Artificial Neural Network (ANN), etc., have been developed in the past few centuries. However, such techniques cannot realize high detection accuracy while using more data and also its training time was high due to the use of more learning variables. Hence, in this article, an effective attribute selection and classification technique is proposed for detecting the phishing attacks in wireless networks. Initially, the email data is collected which comprises more attributes to be extracted by the attribute extractor. Then, the Elephant Herding Optimization (EHO) algorithm is applied to choose the most relevant attributes among the all extracted attributes related to the email. Moreover, the chosen attributes are fed to the Deep Convolutional Neural Network (DCNN)-based classifier for identifying the phishing and authentic emails. Finally, the outcomes show that EHO-DCNN technique achieves a better detection performance when compared to previous techniques.