Deep Boltzmann Machine Based Detection of Phishing URLS
The transparency of the Web and Internet exposes opportunities for criminals to upload malicious content.
Simultaneously with the enhancement of online business exchanges, Phishing - the act of stealing individual information
which rises in number. The phishers endeavor to make fake-sites appear to be like true legitimate sites both in terms of
interface and uniform asset locator (URL) address. Subsequently, the records of casualty have been rising due to ineffectual
methods using the blacklist to recognize phishing. The existing methods make use of all extracted features in the phishing
URL detection, leading to high false positive rate. This paper proposes an automatic phishing identification method using
deep learning approach for detecting unknown URL is either a phishing URL or benign URL. Deep Boltzmann Machine
(DBM) is utilized for pre-training the model with a superior representation of information for feature selection and binary
classification of benign and phishing URL is performed utilizing Deep Neural Network (DNN) which accomplishes higher
recognition rate of phishing URLs with low false positive rate than other machine learning techniques.
Keywords - Phishing, Phishing URL detection, Deep Boltzmann Machine, Deep Neural Network.