arXiv subjects

Juhee Kim

Publications and source records attributed to Juhee Kim.

7 recordsLinked to original sources

arXiv2026

Agent Data Injection Attacks are Realistic Threats to AI Agents

AI agents act on behalf of user prompts, consuming external data and taking actions based on the agent context. Prior research on AI agent security has primarily focused on indirect prompt injection (IPI). Its most well-studied category is instruction injection, where attacker-controlled untrusted data is interpreted as an instruction. In response, many mitigations have been proposed to prevent instruction injection attacks. In this paper, we introduce a new category of IPI, agent data injection attacks (ADI). ADI injects malicious data disguised as trusted data, such as security-critical metadata (e.g., resource identifiers or data origins) or agent context data (e.g., tool call and response formats). As a result, agents unknowingly execute unintended actions based on attacker-controlled data. ADI has similar attack impacts as instruction injection attacks, because it causes agents to misbehave and execute unintended actions. Despite the similar impact, ADI remains underexplored and easily bypasses existing IPI defenses. We found several critical vulnerabilities in real-world agents that allow an attacker to launch various attacks: arbitrary click attacks on web agents (Claude in Chrome, Antigravity, and Nanobrowser), and remote code execution and supply-chain attacks on coding agents (Claude Code, Codex, and Gemini CLI). We evaluate ADI vulnerabilities across off-the-shelf models and AI agents, and find that ADI is effective in both standalone LLMs and AI agent settings. ADI exposes a critical gap in agent security, signifying that current AI agents do not employ a fundamental security principle: current agents do not isolate trusted data from untrusted data.

cs.CR↗

arXiv2026

DualView: Preventing Indirect Prompt Injection in Personal AI Agents

Personal AI agents that run on the user's local machine, such as OpenClaw, automate daily tasks including web search, email, and file management. Their access to computer resources, including the network, file system, and shell, exposes them to indirect prompt injection (IPI) attacks. Prior Dual LLM defenses block IPI by replacing untrusted data with symbols that the agent can reference but not read. However, they track untrusted data only inside the agent's context, so when the agent saves and later rereads untrusted data, that data, possibly an attacker's prompt, can return as trusted data rather than as a symbol, which we call stored IPI. Operating on the user's real environment, which humans and programs share, is what makes agents like OpenClaw practical, and is exactly why a defense that ignores it is incomplete. Preserving symbols in such an environment is hard, because humans and programs need original data. We present DualView, which extends untrusted data tracking from the agent's context to the user's environment, including the file system, shell, network, and other agents, by giving each channel two views. In AgentView, the agent sees untrusted data as symbols even after writing it out and reading it back, blocking stored IPI, while HumanView preserves original data for humans and tools. DualView routes each tool call to the right view and synchronizes data across the two views. DualView deploys as an OpenClaw plugin using only tool hooks, without changing the agent's tool-call logic or tool implementations. Since DualView isolates untrusted data by design, its protection is not limited to known attack templates. In our evaluation on an IPI benchmark and PinchBench, DualView blocked every IPI attack, including stored IPI, while keeping utility close to the unprotected baseline.

cs.CR↗

arXiv2026

The Attack and Defense Landscape of Agentic AI: A Comprehensive Survey

AI agents that combine large language models with non-AI system components are rapidly emerging in real-world applications, offering unprecedented automation and flexibility. However, this unprecedented flexibility introduces complex security challenges fundamentally different from those in traditional software systems. This paper presents the first systematic and comprehensive survey of AI agent security, including an analysis of the design space, attack landscape, and defense mechanisms for secure AI agent systems. We further conduct case studies to point out existing gaps in securing agentic AI systems and identify open challenges in this emerging domain. Our work also introduces the first systematic framework for understanding the security risks and defense strategies of AI agents, serving as a foundation for building both secure agentic systems and advancing research in this critical area.

cs.CR↗

arXiv2026

How to recover a permutation group amidst errors

We consider the problem of recovering a permutation group $G \leq S_n$ from an error-prone sampling process $X$. We model $X$ as an $S_n$-valued random variable, defined as a mixture of the uniform distributions on $G$ and $S_n$ . Our suite of tools recovers properties of $G$ from $X$ and bolsters our main method for recovering $G$ itself. Our algorithms are motivated by the numerical computation of monodromy groups, a setting where such error-prone sampling procedures occur organically.

math.ST↗

arXiv2025

Prompt Flow Integrity to Prevent Privilege Escalation in LLM Agents

Large Language Models (LLMs) are combined with tools to create powerful LLM agents that provide a wide range of services. Unlike traditional software, LLM agent's behavior is determined at runtime by natural language prompts from either user or tool's data. This flexibility enables a new computing paradigm with unlimited capabilities and programmability, but also introduces new security risks, vulnerable to privilege escalation attacks. Moreover, user prompts are prone to be interpreted in an insecure way by LLM agents, creating non-deterministic behaviors that can be exploited by attackers. To address these security risks, we propose Prompt Flow Integrity (PFI), a system security-oriented solution to prevent privilege escalation in LLM agents. Analyzing the architectural characteristics of LLM agents, PFI features three mitigation techniques -- i.e., agent isolation, secure untrusted data processing, and privilege escalation guardrails. Our evaluation result shows that PFI effectively mitigates privilege escalation attacks while successfully preserving the utility of LLM agents.

cs.CR↗

arXiv2025

Dialogue Systems for Emotional Support via Value Reinforcement

Emotional support dialogue systems aim to reduce help-seekers' distress and help them overcome challenges. While human values$\unicode{x2013}$core beliefs that shape an individual's priorities$\unicode{x2013}$are increasingly emphasized in contemporary psychological therapy for their role in fostering internal transformation and long-term emotional well-being, their integration into emotional support systems remains underexplored. To bridge this gap, we present a value-driven method for training emotional support dialogue systems designed to reinforce positive values in seekers. Notably, our model identifies which values to reinforce at each turn and how to do so, by leveraging online support conversations from Reddit. We evaluate the method across support skills, seekers' emotional intensity, and value reinforcement. Our method consistently outperforms various baselines, effectively exploring and eliciting values from seekers. Additionally, leveraging crowd knowledge from Reddit significantly enhances its effectiveness. Therapists highlighted its ability to validate seekers' challenges and emphasize positive aspects of their situations$\unicode{x2013}$both crucial elements of value reinforcement. Our work, being the first to integrate value reinforcement into emotional support systems, demonstrates its promise and establishes a foundation for future research.

cs.CL↗

arXiv2024

TikTag: Breaking ARM's Memory Tagging Extension with Speculative Execution

ARM Memory Tagging Extension (MTE) is a new hardware feature introduced in ARMv8.5-A architecture, aiming to detect memory corruption vulnerabilities. The low overhead of MTE makes it an attractive solution to mitigate memory corruption attacks in modern software systems and is considered the most promising path forward for improving C/C++ software security. This paper explores the potential security risks posed by speculative execution attacks against MTE. Specifically, this paper identifies new TikTag gadgets capable of leaking the MTE tags from arbitrary memory addresses through speculative execution. With TikTag gadgets, attackers can bypass the probabilistic defense of MTE, increasing the attack success rate by close to 100%. We demonstrate that TikTag gadgets can be used to bypass MTE-based mitigations in real-world systems, Google Chrome and the Linux kernel. Experimental results show that TikTag gadgets can successfully leak an MTE tag with a success rate higher than 95% in less than 4 seconds. We further propose new defense mechanisms to mitigate the security risks posed by TikTag gadgets.

cs.CR↗