Geng Hong

Jiaqi Luo, Songyang Peng, Jiarun Dai et al.

LLM-based agents have recently attracted significant attention due to their ability to autonomously invoke relevant tools to accomplish complex tasks. However, recent studies have shown that these agents face severe security risks, which may lead to privacy leakage, financial loss, or even full system compromise. In this paper, we present AgentGuard, an attribute-based access control framework for tool-use LLM-based agents. AgentGuard adopts a client-server architecture. On the client side, AgentGuard provides lightweight integration for agents implemented in different programming languages and architectures. It requires only minor code modifications (e.g., around 10 lines) without changing the underlying agent execution logic. On the server side, AgentGuard provides three complementary inspection mechanisms to cover both single-tool and cross-tool security risks in agent execution. In addition, it offers a visualized front-end interface for security policy specification and runtime auditing. Currently, AgentGuard is publicly accessible at https://github.com/WhitzardAgent/AgentGuard.

Changyue Jiang, Wenqi Zhang, Xudong Pan et al.

LLM-based agents solve complex tasks through iterative reasoning, tool use, and environment interaction, where each intermediate thought directly shapes subsequent actions. Small deviations in these thoughts can therefore propagate into unsafe behaviors, yet existing guardrails typically operate only on final outputs or require intrusive model modifications. We introduce Thought-Aligner, a lightweight plug-in safety model that performs causal correction on unsafe thoughts before action execution, without altering the underlying agent. The corrected thoughts are fed back into the agent, steering its decision process and tool use toward safer trajectories. Because it operates solely at the thought level, Thought-Aligner is model-agnostic and can be integrated into diverse agent frameworks. We train Thought-Aligner via two-stage contrastive learning on paired safe and unsafe thoughts generated across ten risk scenarios. Experiments on diverse agent-safety benchmarks and six LLMs show that Thought-Aligner increases behavioral safety from about 50% without protection to around 90% on average, exceeding state-of-the-art guardrails by roughly 23%, while also improving helpfulness by about 5%. The method incurs low per-step latency and minimal overhead, enabling scalable and practical deployment. We publicly release Thought-Aligner-7B at https://huggingface.co/WhitzardAgent/Thought-Aligner-7B.

Xiang Yang, Feifei Li, Mi Zhang et al.

Recent Text-to-Image (T2I) models based on rectified-flow transformers (e.g., SD3, FLUX) achieve high generative fidelity but remain vulnerable to unsafe semantics, especially when triggered by multi-token interactions. Existing mitigation methods largely rely on fine-tuning or attention modulation for concept unlearning; however, their expensive computational overhead and design tailored to U-Net-based denoisers hinder direct adaptation to transformer-based diffusion models (e.g., MMDiT). In this paper, we conduct an in-depth analysis of the attention mechanism in MMDiT and find that unsafe semantics concentrate within interpretable, low-dimensional subspaces at head level, where a finite set of safety-critical heads is responsible for unsafe feature extraction. We further observe that perturbing the Rotary Positional Embedding (RoPE) applied to the query and key vectors can effectively modify some specific concepts in the generated images. Motivated by these insights, we propose SafeRoPE, a lightweight and fine-grained safe generation framework for MMDiT. Specifically, SafeRoPE first constructs head-wise unsafe subspaces by decomposing unsafe embeddings within safety-critical heads, and computes a Latent Risk Score (LRS) for each input vector via projection onto these subspaces. We then introduce head-wise RoPE perturbations that can suppress unsafe semantics without degrading benign content or image quality. SafeRoPE combines both head-wise LRS and RoPE perturbations to perform risk-specific head-wise rotation on query and key vector embeddings, enabling precise suppression of unsafe outputs while maintaining generation fidelity. Extensive experiments demonstrate that SafeRoPE achieves SOTA performance in balancing effective harmful content mitigation and utility preservation for safe generation of MMDiT. Codes are available at https://github.com/deng12yx/SafeRoPE.

Xinyi Wu, Geng Hong, Yueyue Chen et al.

Web agents, powered by large language models (LLMs), are increasingly deployed to automate complex web interactions. The rise of open-source frameworks (e.g., Browser Use, Skyvern-AI) has accelerated adoption, but also broadened the attack surface. While prior research has focused on model threats such as prompt injection and backdoors, the risks of social engineering remain largely unexplored. We present the first systematic study of social engineering attacks against web automation agents and design a pluggable runtime mitigation solution. On the attack side, we introduce the AgentBait paradigm, which exploits intrinsic weaknesses in agent execution: inducement contexts can distort the agent's reasoning and steer it toward malicious objectives misaligned with the intended task. On the defense side, we propose SUPERVISOR, a lightweight runtime module that enforces environment and intention consistency alignment between webpage context and intended goals to mitigate unsafe operations before execution. Empirical results show that mainstream frameworks are highly vulnerable to AgentBait, with an average attack success rate of 67.5% and peaks above 80% under specific strategies (e.g., trusted identity forgery). Compared with existing lightweight defenses, our module can be seamlessly integrated across different web automation frameworks and reduces attack success rates by up to 78.1% on average while incurring only a 7.7% runtime overhead and preserving usability. This work reveals AgentBait as a critical new threat surface for web agents and establishes a practical, generalizable defense, advancing the security of this rapidly emerging ecosystem. We reported the details of this attack to the framework developers and received acknowledgment before submission.

Yiming Sun, Mi Zhang, Feifei Li et al.

Despite Video Large Language Models having rapidly advanced in recent years, perceptual hallucinations pose a substantial safety risk, which severely restricts their real-world applicability. While several methods for hallucination mitigation have been proposed, they often compromise the model's capacity for video understanding and reasoning. In this work, we propose SmartSight, a pioneering step to address this issue in a training-free manner by leveraging the model's own introspective capabilities. Specifically, SmartSight generates multiple candidate responses to uncover low-hallucinated outputs that are often obscured by standard greedy decoding. It assesses the hallucination of each response using the Temporal Attention Collapse score, which measures whether the model over-focuses on trivial temporal regions of the input video when generating the response. To improve efficiency, SmartSight identifies the Visual Attention Vanishing point, enabling more accurate hallucination estimation and early termination of hallucinated responses, leading to a substantial reduction in decoding cost. Experiments show that SmartSight substantially lowers hallucinations for Qwen2.5-VL-7B by 10.59% on VRIPT-HAL, while simultaneously enhancing video understanding and reasoning, boosting performance on VideoMMMU by up to 8.86%. These results highlight SmartSight's effectiveness in improving the reliability of open-source Video-LLMs.

Wenxuan Li, Zhenfei Zhang, Mi Zhang et al.

Large language models (LLMs) may memorize sensitive or copyrighted content, raising significant privacy and legal concerns. While machine unlearning has emerged as a potential remedy, prevailing paradigms rely on user-provided forget sets, making unlearning requests difficult to audit and exposing systems to secondary leakage and malicious abuse. We propose MAGE, a Memory-grAph Guided Erasure framework for user-minimized, corpus-free unlearning. Given only a lightweight user anchor that identifies a target entity, MAGE probes the target LLM to recover target-related memorization, organizes it into a weighted local memory graph, and synthesizes scoped supervision for unlearning. MAGE is model-agnostic, can be plugged into standard unlearning methods, and requires no access to the original training corpus. Experiments on two benchmarks, TOFU and RWKU, demonstrate that MAGE's self-generated supervision achieves effective unlearning performance comparable to supervision generated with external reference, while preserving overall utility. These results support a practical and auditable unlearning workflow driven by minimal anchors rather than user-supplied forget corpora.

Yuanmin Huang, Mi Zhang, Chen Chen et al.

While diffusion models excel at generating high-quality images, their tendency to memorize training data poses significant privacy and copyright risks. In this work, we for the first time identify that memorization induces internal numerical instability, often manifesting as visually ``broken'' artifacts. Inspired by stability analysis in numerical methods, we introduce empirical stability regions based on latent update norms to quantitatively characterize stable behavior during generation. Leveraging this, we propose a principled, on-the-fly framework for step-wise detection and adaptive mitigation. Our approach suppresses memorization without altering prompts or guidance, thereby preserving semantic fidelity and image quality. Extensive experiments on Stable Diffusion 1.4 demonstrate that our method achieves an AUC $>0.999$ detection performance and a $0.0\%$ memorization rate after mitigation with negligible overhead ($\approx0.01$s per image).

Pei Chen, Geng Hong, Xinyi Wu et al.

The emergence of Large Language Model-enhanced Search Engines (LLMSEs) has revolutionized information retrieval by integrating web-scale search capabilities with AI-powered summarization. While these systems demonstrate improved efficiency over traditional search engines, their security implications against well-established black-hat Search Engine Optimization (SEO) attacks remain unexplored. In this paper, we present the first systematic study of SEO attacks targeting LLMSEs. Specifically, we examine ten representative LLMSE products (e.g., ChatGPT, Gemini) and construct SEO-Bench, a benchmark comprising 1,000 real-world black-hat SEO websites, to evaluate both open- and closed-source LLMSEs. Our measurements show that LLMSEs mitigate over 99.78% of traditional SEO attacks, with the phase of retrieval serving as the primary filter, intercepting the vast majority of malicious queries. We further propose and evaluate seven LLMSEO attack strategies, demonstrating that off-the-shelf LLMSEs are vulnerable to LLMSEO attacks, i.e., rewritten-query stuffing and segmented texts double the manipulation rate compared to the baseline. This work offers the first in-depth security analysis of the LLMSE ecosystem, providing practical insights for building more resilient AI-driven search systems. We have responsibly reported the identified issues to major vendors.

Yulin Shen, Xudong Pan, Geng Hong et al.

Recent advances in the Model Context Protocol (MCP) have enabled large language models (LLMs) to invoke external tools with unprecedented ease. This creates a new class of powerful and tool augmented agents. Unfortunately, this capability also introduces an under explored attack surface, specifically the malicious manipulation of tool responses. Existing techniques for indirect prompt injection that target MCP suffer from high deployment costs, weak semantic coherence, or heavy white box requirements. Furthermore, they are often easily detected by recently proposed defenses. In this paper, we propose Tree structured Injection for Payloads (TIP), a novel black-box attack which generates natural payloads to reliably seize control of MCP enabled agents even under defense. Technically, We cast payload generation as a tree structured search problem and guide the search with an attacker LLM operating under our proposed coarse-to-fine optimization framework. To stabilize learning and avoid local optima, we introduce a path-aware feedback mechanism that surfaces only high quality historical trajectories to the attacker model. The framework is further hardened against defensive transformations by explicitly conditioning the search on observable defense signals and dynamically reallocating the exploration budget. Extensive experiments on four mainstream LLMs show that TIP attains over 95% attack success in undefended settings while requiring an order of magnitude fewer queries than prior adaptive attacks. Against four representative defense approaches, TIP preserves more than 50% effectiveness and significantly outperforms the state-of-the-art attacks. By implementing the attack on real world MCP systems, our results expose an invisible but practical threat vector in MCP deployments. We also discuss potential mitigation approaches to address this critical security gap.

Xinyi Wu, Jiagui Chen, Geng Hong et al.

Web Agents are increasingly deployed to perform complex tasks in real web environments, yet their security evaluation remains fragmented and difficult to standardize. We present WebTrap Park, an automated platform for systematic security evaluation of Web Agents through direct observation of their concrete interactions with live web pages. WebTrap Park instantiates three major sources of security risk into 1,226 executable evaluation tasks and enables action based assessment without requiring agent modification. Our results reveal clear security differences across agent frameworks, highlighting the importance of agent architecture beyond the underlying model. WebTrap Park is publicly accessible at https://security.fudan.edu.cn/webagent and provides a scalable foundation for reproducible Web Agent security evaluation.

Changyi Li, Jiayi Wang, Xudong Pan et al.

Large Reasoning Models (LRMs) leverage transparent reasoning traces, known as Chain-of-Thoughts (CoTs), to break down complex problems into intermediate steps and derive final answers. However, these reasoning traces introduce unique safety challenges: harmful content can be embedded in intermediate steps even when final answers appear benign. Existing moderation tools, designed to handle generated answers, struggle to effectively detect hidden risks within CoTs. To address these challenges, we introduce ReasoningShield, a lightweight yet robust framework for moderating CoTs in LRMs. Our key contributions include: (1) formalizing the task of CoT moderation with a multi-level taxonomy of 10 risk categories across 3 safety levels, (2) creating the first CoT moderation benchmark which contains 9.2K pairs of queries and reasoning traces, including a 7K-sample training set annotated via a human-AI framework and a rigorously curated 2.2K human-annotated test set, and (3) developing a two-stage training strategy that combines stepwise risk analysis and contrastive learning to enhance robustness. Experiments show that ReasoningShield achieves state-of-the-art performance, outperforming task-specific tools like LlamaGuard-4 by 35.6% and general-purpose commercial models like GPT-4o by 15.8% on benchmarks, while also generalizing effectively across diverse reasoning paradigms, tasks, and unseen scenarios. All resources are released at https://github.com/CosmosYi/ReasoningShield.

Yichen Wu, Xudong Pan, Geng Hong et al.

As the general capabilities of large language models (LLMs) improve and agent applications become more widespread, the underlying deception risks urgently require systematic evaluation and effective oversight. Unlike existing evaluation which uses simulated games or presents limited choices, we introduce OpenDeception, a novel deception evaluation framework with an open-ended scenario dataset. OpenDeception jointly evaluates both the deception intention and capabilities of LLM-based agents by inspecting their internal reasoning process. Specifically, we construct five types of common use cases where LLMs intensively interact with the user, each consisting of ten diverse, concrete scenarios from the real world. To avoid ethical concerns and costs of high-risk deceptive interactions with human testers, we propose to simulate the multi-turn dialogue via agent simulation. Extensive evaluation of eleven mainstream LLMs on OpenDeception highlights the urgent need to address deception risks and security concerns in LLM-based agents: the deception intention ratio across the models exceeds 80%, while the deception success rate surpasses 50%. Furthermore, we observe that LLMs with stronger capabilities do exhibit a higher risk of deception, which calls for more alignment efforts on inhibiting deceptive behaviors.

Wenqi Zhang, Yulin Shen, Changyue Jiang et al.

Large foundation models are integrated into Computer Use Agents (CUAs), enabling autonomous interaction with operating systems through graphical user interfaces (GUIs) to perform complex tasks. This autonomy introduces serious security risks: malicious instructions or visual prompt injections can trigger unsafe reasoning and cause harmful system-level actions. Existing defenses, such as detection-based blocking, prevent damage but often abort tasks prematurely, reducing agent utility. In this paper, we present MirrorGuard, a plug-and-play defense framework that uses simulation-based training to improve CUA security in the real world. To reduce the cost of large-scale training in operating systems, we propose a novel neural-symbolic simulation pipeline, which generates realistic, high-risk GUI interaction trajectories entirely in a text-based simulated environment, which captures unsafe reasoning patterns and potential system hazards without executing real operations. In the simulation environment, MirrorGuard learns to intercept and rectify insecure reasoning chains of CUAs before they produce and execute unsafe actions. In real-world testing, extensive evaluations across diverse benchmarks and CUA architectures show that MirrorGuard significantly mitigates security risks. For instance, on the ByteDance UI-TARS system, it reduces the unsafe rate from 66.5% to 13.0% while maintaining a marginal false refusal rate (FRR). In contrast, the state-of-the-art GuardAgent only achieves a reduction to 53.9% and suffers from a 15.4% higher FRR. Our work proves that simulation-derived defenses can provide robust, real-world protection while maintaining the fundamental utility of the agent. Our code and model are publicly available at https://bmz-q-q.github.io/MirrorGuard/.

Yining Wang, Yuquan Wang, Xi Li et al.

As Large Language Models (LLMs) continue to advance, they are increasingly relied upon as real-time sources of information by non-expert users. To ensure the factuality of the information they provide, much research has focused on mitigating hallucinations in LLM responses, but only in the context of formal user queries, rather than maliciously crafted ones. In this study, we introduce The Illusionist's Prompt, a novel hallucination attack that incorporates linguistic nuances into adversarial queries, challenging the factual accuracy of LLMs against five types of fact-enhancing strategies. Our attack automatically generates highly transferrable illusory prompts to induce internal factual errors, all while preserving user intent and semantics. Extensive experiments confirm the effectiveness of our attack in compromising black-box LLMs, including commercial APIs like GPT-4o and Gemini-2.0, even with various defensive mechanisms.

Yuquan Wang, Mi Zhang, Yining Wang et al.

Large Reasoning Models (LRMs) have demonstrated impressive performance in reasoning-intensive tasks, but they remain vulnerable to harmful content generation, particularly in the mid-to-late steps of their reasoning processes. Existing defense mechanisms, however, rely on costly fine-tuning and additional expert knowledge, which restricts their scalability. In this work, we propose ReasoningGuard, an inference-time safeguard for LRMs, which injects timely safety aha moments to steer harmless while helpful reasoning processes. Leveraging the model's internal attention behavior, our approach accurately identifies critical points in the reasoning path, and triggers spontaneous, safety-oriented reflection. To safeguard both the subsequent reasoning steps and the final answers, we further implement a scaling sampling strategy during the decoding phase, selecting the optimal reasoning path. Inducing minimal extra inference cost, ReasoningGuard effectively mitigates three types of jailbreak attacks, including the latest ones targeting the reasoning process of LRMs. Our approach outperforms seven existing safeguards, achieving state-of-the-art safety defenses while effectively avoiding the common exaggerated safety issues.

Xinyi Wu, Geng Hong, Pei Chen et al.

As large language models (LLMs) advance, concerns about their misconduct in complex social contexts intensify. Existing research overlooked the systematic understanding and assessment of their criminal capability in realistic interactions. We propose a unified framework PRISON, to quantify LLMs' criminal potential across five traits: False Statements, Frame-Up, Psychological Manipulation, Emotional Disguise, and Moral Disengagement. Using structured crime scenarios adapted from classic films grounded in reality, we evaluate both criminal potential and anti-crime ability of LLMs. Results show that state-of-the-art LLMs frequently exhibit emergent criminal tendencies, such as proposing misleading statements or evasion tactics, even without explicit instructions. Moreover, when placed in a detective role, models recognize deceptive behavior with only 44% accuracy on average, revealing a striking mismatch between conducting and detecting criminal behavior. These findings underscore the urgent need for adversarial robustness, behavioral alignment, and safety mechanisms before broader LLM deployment.