Abstract: Security issues in unmanned systems have become a key constraint on their reliable deployment. This paper focuses on security threats to autonomous unmanned systems in embodied intelligence scenarios. It systematically categorizes major attack types across different layers, with particular emphasis on two primary forms of attacks in perception systems—contact-based and non-contact-based. Furthermore, it examines the vulnerable modes and potential vulnerabilities of embodied intelligent systems during task execution. By analyzing representative attack methods and defense strategies from existing research, this paper provides insights for enhancing the security and robustness of autonomous unmanned systems.