Regresar

Analysis and Design of a Pbkp_redictive Model for Phishing Detection Based on Url and Email Corpus

Abstract:

One of the most reported cyber crimes worldwide is phishing, and various anti-phishing systems (APS) are currently being developed to identify this type of attack on communication systems in real time. Despite the efforts of organizations, this attack continues to grow, due to the erroneous detection in the zero-day attack: the high computational cost and the high rates of forgery. Although the Machine Learning (ML) approach has achieved a favorable accuracy rate, it should be considered that the choice and performance of the feature vector is a key point to obtain an adequate level of accuracy. In this work, a pbkp_redictive model based on ML and the analysis of the efficiency of some anti-phishing schemes that served to understand this issue is proposed. The proposed model consists of a feature selection module that is used to build the final vector. These characteristics are extracted from the URL, the properties of the web page, and the email corpus. The system uses the Random Forest (RF) and Naïve Bayes (NB) classification models, which have been trained on the feature vector. The experiments were based on datasets composed of phishing and benign instances. Using cross-validation, the experimental results indicate a precision of 97.5% for the datasets used, while a precision of 96.5% was obtained for the approach of this research at the local level.