mlsanityData to 14 Oct 2026

Security and Privacy in Large-Model-Driven Embodied Agents: Attacks, Defenses, and Future Directions

arXiv 2609.27847 ↗First posted 19 Aug 2026cs.CRcited by 0

On the map: LLM agents and tool use → Agent threats, guardrails and governance

Abstract (arXiv)

Large-model-driven embodied agents integrate foundation models with perception, reasoning, planning, and physical action, extending conventional model-level risks into embodied closed loops. Existing studies on their security and privacy remain fragmented across different system components and operational stages, making it difficult to understand how risks arise, propagate, and ultimately affect physical behavior or sensitive information. This survey presents a lifecycle-based analysis of security and privacy in large-model-driven embodied agents. We organize existing research into five stages: model construction and supply chain, multimodal input and interaction, semantic reasoning and task planning, action execution and physical feedback, and long-term deployment. Within this lifecycle, we systematically review representative attacks, defenses, and evaluation methods. Our analysis shows that attack entry, consequence realization, and defense intervention often occur at different stages of the embodied closed loop. It further reveals substantial gaps in end-to-end protection, real-world evaluation, and long-term privacy governance. This survey provides a unified perspective for understanding current progress and identifying critical directions for securing large-model-driven embodied agents.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use. Builds on 87 of its 120 references that are in the map. First posted on arXiv 19 Aug 2026.

Builds on

87 of 120 references.

Closest ancestors · 14

Same direction, newest first

  1. 2025RoboSafe: Safeguarding Embodied Agents via Executable Safety Logic24 Dec
  2. 2025MADRA: Multi-Agent Debate for Risk-Aware Embodied Planning26 Nov
  3. 2025SENTINEL: A Multi-Level Formal Framework for Safety Evaluation of Foundation Model-based Embodied Agents14 Oct
  4. 2025SafeMind: Benchmarking and Mitigating Safety Risks in Embodied LLM Agents30 Sep
  5. 2025ADVEDM:Fine-grained Adversarial Attack against VLM-based Embodied Agents20 Sep
  6. 2025IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks19 Jun
  7. 2025AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions17 Jun
  8. 2025Advancing Embodied Agent Security: From Safety Benchmarks to Input Moderation22 Apr
  9. 2025A Framework for Benchmarking and Aligning Task-Planning Safety in LLM-Based Embodied Agents20 Apr
  10. 2024SafeAgentBench: A Benchmark for Safe Task Planning of Embodied LLM Agents17 Dec
  11. 2024Coherence-Driven Multimodal Safety Dialogue with Active Learning for Embodied Agents18 Oct
  12. 2024EARBench: Towards Evaluating Physical Risk Awareness for Task Planning of Foundation Model-based Embodied AI Agents8 Aug
2 more
  1. 2024BadRobot: Jailbreaking Embodied LLM Agents in the Physical World16 Jul
  2. 2023Plug in the Safety Chip: Enforcing Constraints for LLM-driven Robot Agents18 Sep

Other references · 73

  1. 2024A Survey on Vision-Language-Action Models for Embodied AIcited by 279
  2. 2023Large Language Models for Robotics: A Surveycited by 133
  3. 2025Are VLMs Ready for Autonomous Driving? An Empirical Study from the Reliability, Data, and Metric Perspectivescited by 122
  4. 2024Exploring the Adversarial Vulnerabilities of Vision-Language-Action Models in Roboticscited by 73
  5. 2025SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learningcited by 72
  6. 2025BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimizationcited by 38
  7. 2025Safety Guardrails for LLM-Enabled Robotscited by 35
  8. 2025Towards Robust and Secure Embodied AI: A Survey on Vulnerabilities and Attackscited by 34
  9. 2025VLSA: Vision-Language-Action Models with Plug-and-Play Safety Constraint Layercited by 34
  10. 2025Generating Robot Constitutions & Benchmarks for Semantic Safetycited by 33
  11. 2024HAZARD Challenge: Embodied Decision Making in Dynamically Changing Environmentscited by 32
  12. 2025Adversarial Attacks on Robotic Vision Language Action Modelscited by 32
38 more
  1. 2025Safety Aware Task Planning via Large Language Models in Roboticscited by 24
  2. 2025Eva-VLA: Evaluating Vision-Language-Action Models' Robustness Under Real-World Physical Variationscited by 24
  3. 2025VLA-AN: An Efficient and Onboard Vision-Language-Action Framework for Aerial Navigation in Complex Environmentscited by 23
  4. 2024LLM-Driven Robots Risk Enacting Discrimination, Violence, and Unlawful Actionscited by 22
  5. 2025Counterfactual VLA: Self-Reflective Vision-Language-Action Model with Adaptive Reasoningcited by 22
  6. 2024Embodied Red Teaming for Auditing Robotic Foundation Modelscited by 21
  7. 2025FreezeVLA: Action-Freezing Attacks against Vision-Language-Action Modelscited by 21
  8. 2025Subtle Risks, Critical Failures: A Framework for Diagnosing Physical Safety of LLMs for Embodied Decision Makingcited by 20
  9. 2025When Alignment Fails: Multimodal Adversarial Attacks on Vision-Language-Action Modelscited by 20
  10. 2024Embodied AI with Two Arms: Zero-shot Learning, Safety and Modularitycited by 19
  11. 2025SoK: The Privacy Paradox of Large Language Models: Advancements, Privacy Risks, and Mitigationcited by 19
  12. 2025Goal-oriented Backdoor Attack against Vision-Language-Action Models via Physical Objectscited by 17
  13. 2024On the Vulnerability of LLM/VLM-Controlled Roboticscited by 15
  14. 2024A Study on Prompt Injection Attack Against LLM-Integrated Mobile Robotic Systemscited by 15
  15. 2024Robot Collapse: Supply Chain Backdoor Attacks Against VLM-based Robotic Manipulationcited by 15
  16. 2024SafeEmbodAI: a Safety Framework for Mobile Robots in Embodied AI Systemscited by 13
  17. 2025MinD: Learning A Dual-System World Model for Real-Time Planning and Implicit Risk Analysiscited by 13
  18. 2025When Robots Obey the Patch: Universal Transferable Patch Attacks on Vision-Language-Action Modelscited by 13
  19. 2026Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanismscited by 13
  20. 2025LLM-Enabled In-Context Learning for Data Collection Scheduling in UAV-assisted Sensor Networkscited by 12
  21. 2024TrustNavGPT: Modeling Uncertainty to Improve Trustworthiness of Audio-Guided LLM-Based Robot Navigationcited by 11
  22. 2025CHAI: Command Hijacking against embodied AIcited by 11
  23. 2026VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Modelscited by 10
  24. 2025Foundation Model Driven Robotics: A Comprehensive Reviewcited by 9
  25. 2025RoboChemist: Long-Horizon and Safety-Compliant Robotic Chemical Experimentationcited by 9
  26. 2025BEAT: Visual Backdoor Attacks on VLM-based Embodied Agents via Contrastive Trigger Learningcited by 9
  27. 2025Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attackscited by 9
  28. 2026State Backdoor: Towards Stealthy Real-world Poisoning Attack on Vision-Language-Action Model in State Spacecited by 9
  29. 2024Hierarchical LLMs In-the-Loop Optimization for Real-Time Multi-Robot Target Tracking under Unknown Hazardscited by 8
  30. 2026Learning from Mistakes: Post-Training for Driving VLA with Takeover Datacited by 8
  31. 2024Safety Control of Service Robots with LLMs and Embodied Knowledge Graphscited by 7
  32. 2026Shifting Uncertainty to Critical Moments: Towards Reliable Uncertainty Quantification for VLA Modelcited by 7
  33. 2026Safety in Embodied AI: A Survey of Risks, Attacks, and Defensescited by 7
  34. 2025The Safety Challenge of World Models for Embodied AI Agents: A Reviewcited by 6
  35. 2025From Words to Safety: Language-Conditioned Safety Filtering for Robot Navigationcited by 6
  36. 2026What Breaks Embodied AI Security:LLM Vulnerabilities, CPS Flaws,or Something Else?cited by 6
  37. 2026SafeManip: A Property-Driven Benchmark for Temporal Safety Evaluation in Robotic Manipulationcited by 6
  38. 2024SwarmGPT: Combining Large Language Models with Safe Motion Planning for Drone Swarm Choreographycited by 5

50 most cited of 73 All 73 (JSON)

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Security of Foundation-Model-Powered Embodied Agents: Attack Surfaces, Attacks, Defenses, and Evaluation
  2. 2026Security of World-Model-Based Embodied AI: A Lifecycle of Threats, Defenses, and Evaluation
  3. 2024Can We Trust Embodied Agents? Exploring Backdoor Attacks against Embodied LLM-based Decision-Making Systems
  4. 2025Position: Human-Robot Interaction in Embodied Intelligence Demands a Shift From Static Privacy Controls to Dynamic Learning ↗
  5. 2025Measuring Physical-World Privacy Awareness of Large Language Models: An Evaluation Benchmark
  6. 2026Benchmarking Jailbreak Guardrails for Embodied Agents
  7. 2025Beyond Model Jailbreak: Systematic Dissection of the "Ten DeadlySins" in Embodied Intelligence
  8. 2025A Survey on Autonomy-Induced Security Risks in Large Model-Based Agents