This comprehensive survey examines the unique security challenges faced by embodied AI systems that interact with the physical world.

• Vulnerability mapping: Categorizes threats across sensor spoofing, adversarial attacks, and failures in task/motion planning
• Attack vectors: Analyzes how physical embodiment creates novel security challenges compared to traditional AI systems
• Defense strategies: Presents frameworks and approaches for enhancing robustness against both environmental and system-level threats
• Future directions: Identifies research gaps and emerging concerns as embodied AI becomes more prevalent

This research is critical for security professionals as embodied AI systems (robots, autonomous vehicles) increasingly operate in sensitive environments where failures could have serious safety and privacy implications.

Towards Robust and Secure Embodied AI: A Survey on Vulnerabilities and Attacks