Leveraging Large Language Models for Automated Detection of Cookie and Session Management Vulnerabilities
Year: 2025
Type: conference-paper
Abstract: In this paper, we explore how Large Language Models (LLMs) can analyze and detect security vulnerabilities in cookies and session management within cybersecurity.Web applications depend on these mechanisms, often leading to risks like session hijacking, XSS, and CSRF, necessitating a thorough understanding as architectures grow complex.This study employ a comparative analysis framework, using datasets of cookie security and cybersecurity logs, and apply prompt engineering to evaluate LLMs in identifying flaws in HTTP headers, analyzing security attributes.Our findings show LLMs can detect insecure cookie configurations, automate assessments, and provide actionable insights, though challenges like false positives, adversarial manipulation, and interpretability persist.This suggests LLMs can enhance security audits and proactive threat mitigation, strengthening web application security.The study highlights some potential of LLMs in automated security analysis at the AI-cybersecurity intersection, with future work aimed at integrating these insights with existing defenses for a more robust, adaptive security framework.
DOI: 10.18260/1-2--55022
Authors: Shashi Chandrappa, Sidike Paheding, Yu Cai
Institutions: Fairfield University, Michigan Technological University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4409905139