Detection of Malicious URLs using Logistic Regression
Usha Sri B, Varun Singh Bamla, Venu Bandari, Manikanta Bheemagani, Chaitanya Uppuganti, Ruthvik Akkenapally · 2024
The rise of mobile devices has driven real- world activities online, exposing users to the growing threat of malicious URLs that can compromise network security by distributing malware, launching phishing attacks, and causing data breaches. This paper proposes using logistic regression, a powerful binary classification tool, to differentiate between malicious and benign URLs. By extracting and analyzing features such as character sequences, suspicious keywords, and specific subdomains, the logistic regression model can identify patterns indicative of malicious intent. Although effective, logistic regression's assumption of linearity between features and maliciousness poses a limitation, as cybercriminals employ complex strategies that may avoid detection. Therefore, while logistic regression offers interpretability and initial success, enhancing this approach with advanced machine learning techniques could provide deeper insights and more robust protection against cyber threats.