Bibliographic citations
Cabezas, A., Gutierrez, J. (2023). Detección de phishing en correos electrónicos mediante procesamiento de lenguaje natural del contenido y URLs ofuscadas [Universidad de Lima]. https://hdl.handle.net/20.500.12724/20761
Cabezas, A., Gutierrez, J. Detección de phishing en correos electrónicos mediante procesamiento de lenguaje natural del contenido y URLs ofuscadas []. PE: Universidad de Lima; 2023. https://hdl.handle.net/20.500.12724/20761
@misc{renati/235756,
title = "Detección de phishing en correos electrónicos mediante procesamiento de lenguaje natural del contenido y URLs ofuscadas",
author = "Gutierrez Sisniegas, Jerome Zack Ricardo",
publisher = "Universidad de Lima",
year = "2023"
}
Phishing is a common type of computer fraud that, through methods like emails or similar messages, requests the input of personal information or access to malicious links. Faced with this issue, various authors have developed email phishing detection models based on content analysis that have demonstrated high detection rates. However, cybercriminals employ new phishing techniques such as the use of obfuscated URLs, which involve modifying malicious URLs to appear legitimate through various methods, such as URL shortening. The present research aims to develop two general methods for detecting phishing in emails using natural language processing. The first method focuses on content analysis, while the second concentrates on analyzing obfuscated URLs. Likewise, the objective is to determine which method enhances the phishing detection rate. Both models involve a preprocessing phase that extracts a set of attributes related to email content and applies natural language processing techniques to the email content. Three datasets, namely "Enron," "Spam Archive," and "Ebbu 2017," were used for model training. To validate the models, metrics such as precision, sensitivity, accuracy, and the area under the curve (AUC) score were considered. The implemented models achieved a maximum precision of 97.02% and 98.70%. Additionally, it was observed that the proposed models yielded results with minimal differences in detecting phishing in emails.
This item is licensed under a Creative Commons License