Abstract
This study proposes a novel, multi-layered approach to adversarial attack detection in machine learning models specifically designed for cybersecurity applications. With the increasing deployment of AI in critical domains such as finance and digital communication, the vulnerability of these systems to adversarial inputs poses a serious threat. The research incorporates a hybrid framework that integrates adversarial detection mechanisms, defense integration levels, and model complexity to improve detection accuracy while reducing false positives. Data were collected from 205 New York-based households and analyzed using both R Studio and SPSS. The findings demonstrate that the proposed model significantly enhances the robustness of cybersecurity systems, offering both technical innovation and practical relevance. This study contributes to the growing body of knowledge on adversarial machine learning and its real-world application in strengthening AI-enabled defense systems, particularly in the U.S. context.