Інтеграція LLM у технологію автоматизованого тестування безпеки веб-застосунків

Abstract

The paper examines the integration specifics of multimodal large language models into automated web application security testing technologies. It proposes an architecture of a multimodal agent-based system, identifying key advantages over unimodal approaches. Main integration risks and their implications are analyzed, with measures to mitigate potential threats outlined. The approach demonstrates considerable promise when properly managing security risks and ensuring effective control of model operations.