Deep Auto-Encoder Neural Network for Phishing Website Classification

In this paper, deep auto-encoder technique proposed for website phishing classification problem. The dataset obtained from UCI which contain most common machine learning datasets. The obtained dataset consists from 30 attributes and the 31th attribute represented (target) there is phishing or not). The first auto-encoder extracted sensitive features and reduce the dimension of features. The second auto-encoder also extracted features from output of first auto-encoder and reduce the dimension of features too. The extracted features classified by using SoftMax classifier. All these parts stacked and trained in supervised learning. The experimental results show that proposed method presented best results than previous works.