Year: 2024
Type: article
Abstract: Phishing attacks are one of the most aggressive vulnerabilities in cybersecurity networks, typically carried out through social engineering and URL obfuscation. Traditional detection methods struggle to combat advanced techniques applied. In this paper, a deep learning-based approach is proposed to increase the accuracy of phishing detection while reducing the number of false positives. Four models: CNN-BLSTM, SNN, Transformer, and DBN, are developed and evaluated on a phishing dataset that includes critical features such as URL structure, domain age, and presence of HTTPS. The other model, CNN-BLSTM, achieved 98.9% better accuracy, effectively linking URL sequences in space and time. It is found that although deep learning models have a significant improvement over traditional methods in detecting phishing attacks, the level of computational resources still prevents them from real-time applications. Further research includes hybrid models and adversarial approaches to improve state-ofthe-art and practical solutions to address phishing threats. This study highlights a new technological application to Internet security concerns, particularly in the area of combating phishing.
Institution: Taif University
Language: English
Topic: Spam and Phishing Detection
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Funder: Taif University
OpenAlex ID: W4405426812