The rapid development of AI has greatly promoted and accelerated transformation of industries, and has been widely applied in fields such as object detection, image classification, speech recognition and etc. However, AI models may have potential security vulnerabilities, which could lead to accidents in practical applications. In this paper, we focused on security of object detection models, and conducted a series of comparative testing based on adversarial examples for typical object detection models, as well as comparative testing of several typical security enhancement methods. The experimental results show that security level of object detection models various against different attack methods, and there is no absolute good models or bad models. As for security enhancement methods, most were designed for one or a certain type of attack method, which can only protect model for certain types of attacks.