AuthGuard-R: Safety-Compliant Mission Hijacking and Dual-Gate Defense for LLM-Controlled Robots

Large Language Models (LLMs) are increasingly used in embodied AI for robotic manipulation and planning, but their deployment in physical systems exposes vulnerabilities such as jailbreak attacks that induce unsafe behavior. Existing inference-time defenses are mainly evaluated on short instructions containing a single hazard, whereas realistic tasks may interleave safe and unsafe sub-tasks, conceal hazardous steps, or combine individually safe actions into unsafe outcomes. Such compound instructions make it harder for a single safety decision to distinguish hazardous sub-tasks from legitimate ones, risking either overlooking the hazard or rejecting the whole instruction. We introduce safe task completion, a task that requires an LLM-enabled robot to refuse the harmful part of an instruction while completing its legitimate part. To address this problem, we propose RuleGraph, a multi-agent Graph-of-Rules planning framework that distributes safety enforcement across multiple reasoning and verification stages. Five cooperating agents are used to ground general safety principles into world-specific rules, assess the instruction's sub-tasks in context, and derive a safe goal from the legitimate ones. Planning toward this goal is then organized as a graph of rule-specific nodes, where higher-level nodes inherit the union of their parents' rules, and candidate plans are progressively verified and aggregated until a single plan satisfies every relevant rule. In this way, RuleGraph reduces reliance on a single centralized safety decision. We evaluate RuleGraph on pure-attack and compound instructions across two benchmarks in two environments using Attack Success Rate (ASR) and Task Success Rate (TSR). RuleGraph reduces ASR to 0% on both benchmarks while achieving the highest TSR among all methods, reaching 98.57% on the RoboGuard benchmark and 84.54% on our long-horizon compound instructions.

Robots need task planning methods to achieve goals that require more than one action. Recently, large pretrained models have demonstrated impressive performance in task planning. For instance, large language models (LLMs) can generate task plans using action and goal descriptions. Despite the rapid progress of large models in robot intelligence, their security implications remain only partially understood, leaving important gaps in the exploration of potential vulnerabilities in LLM-driven robotic planning systems. To investigate such risks, in this paper, we develop MuTRAP, the first multi-trigger trojan attack specifically designed and targeted for LLM-assisted robot task planners. MuTRAP follows the standard practice of LLM usage in robotics where the backbone LLM is typically frozen and hosted in a central server limiting attacker's reach. In contrast, MuTRAP injects backdoor using a small set of task-specific parameters. In addition, we develop a trigger optimization method for selecting multiple-trigger words that are most effective for different robot applications. For instance, one can use unique trigger word "herical" to activate a specific malicious behavior, e.g., cutting hand on a kitchen robot. Through MuTRAP that demonstrates the vulnerability of current LLM-based planners, our goal is to promote the development of secured robot intelligence. Details and demos are provided in: https://mutrap.github.io/MuTRAP/

Advances in advanced artificial intelligence tools have sparked research in robot autonomy, but the development of such systems has largely focused on execution rather than verifying the feasibility actions planning models propose. Like general-purpose LLMs, robotics planning models carry risks: biased toward user-specified goals, they may suggest actions misaligned with scientific ethics, they may be unsafe due to an inability to "remember" prior safety risks, or they may be vulnerable to adversarial attacks on the autonomy ecosystem. We propose a LLM-driven verification layer between planning and execution to evaluate action permissibility. Our LLM-as-a-Judge ensemble combines chain-of-thought reasoning across models and synthesizes those expert judge outputs, mirroring a combination of a mixture of experts and self-consistency approach. This layer serves as middleware, gating plans from the server's planning module before they reach the MCP server and therefore the robot's low-level controls: plans are approved, rejected for reformulation, or escalated for human review. With this system, we achieve near 85% precision across accept/escalate/reject categories 97% containment of adversarial attacks, with negligible errors between accepting and rejecting tasks, and errors mostly manifesting at the escalate boundary.