As the prevalence of intelligent systems continues to grow, attention to their security has intensified. Unlike traditional software testing, security assessments for intelligent systems need encompass models, frameworks, and data to provide a comprehensive understanding of system security. This paper introduces a novel security testing methodology tailored for intelligent systems, capable of accurately evaluating the security of their constituent components and providing a synthesis score. A six-step process for security testing, including system modeling, threat scenario construction, test configuration, test case generation, task scheduling, results evaluation, is outlined. Additionally, a prototype tool has been implemented to conduct security testing across various intelligent systems. The effectiveness of the proposed methodology is demonstrated through the testing of a smart car object, resulting in a synthesis score of 31.86 points. Radar charts are utilized to visualize the test results across different dimensions of the smart car object. Furthermore, limitations of the current methodology are discussed, particularly concerning model security and data privacy.