Large language models (LLMs) are increasingly deployed on edge devices under strict computation and quantization constraints, yet their security implications remain unclear. We study query-based knowledge extraction from quantized edge-deployed LLMs under realistic query budgets and show that, although quantization introduces noise, it does not remove the underlying semantic knowledge, allowing substantial behavioral recovery through carefully designed queries. To systematically analyze this risk, we propose \textbf{CLIQ} (\textbf{Cl}ustered \textbf{I}nstruction \textbf{Q}uerying), a structured query construction framework that improves semantic coverage while reducing redundancy. Experiments on quantized Qwen models (INT8/INT4) demonstrate that CLIQ consistently outperforms original queries across BERTScore, BLEU, and ROUGE, enabling more efficient extraction under limited budgets. These results indicate that quantization alone does not provide effective protection against query-based extraction, highlighting a previously underexplored security risk in edge-deployed LLMs.

As generative AI achieves hyper-realism, superficial artifact detection has become obsolete. While prevailing methods rely on resource-intensive fine-tuning of black-box backbones, we propose that forgery detection capability is already encoded within pre-trained models rather than requiring end-to-end retraining. To elicit this intrinsic capability, we propose the discriminative neural anchors (DNA) framework, which employs a coarse-to-fine excavation mechanism. First, by analyzing feature decoupling and attention distribution shifts, we pinpoint critical intermediate layers where the focus of the model logically transitions from global semantics to local anomalies. Subsequently, we introduce a triadic fusion scoring metric paired with a curvature-truncation strategy to strip away semantic redundancy, precisely isolating the forgery-discriminative units (FDUs) inherently imprinted with sensitivity to forgery traces. Moreover, we introduce HIFI-Gen, a high-fidelity synthetic benchmark built upon the very latest models, to address the lag in existing datasets. Experiments demonstrate that by solely relying on these anchors, DNA achieves superior detection performance even under few-shot conditions. Furthermore, it exhibits remarkable robustness across diverse architectures and against unseen generative models, validating that waking up latent neurons is more effective than extensive fine-tuning.

While Large Language Models (LLMs) have powerful capabilities, they remain vulnerable to jailbreak attacks, which is a critical barrier to their safe web real-time application. Current commercial LLM providers deploy output guardrails to filter harmful outputs, yet these defenses are not impenetrable. Due to LLMs' reliance on autoregressive, token-by-token inference, their semantic representations lack robustness to spatially structured perturbations, such as redistributing tokens across different rows, columns, or diagonals. Exploiting the Transformer's spatial weakness, we propose SpatialJB to disrupt the model's output generation process, allowing harmful content to bypass guardrails without detection. Comprehensive experiments conducted on leading LLMs get nearly 100% ASR, demonstrating the high effectiveness of SpatialJB. Even after adding advanced output guardrails, like the OpenAI Moderation API, SpatialJB consistently maintains a success rate exceeding 75%, outperforming current jailbreak techniques by a significant margin. The proposal of SpatialJB exposes a key weakness in current guardrails and emphasizes the importance of spatial semantics, offering new insights to advance LLM safety research. To prevent potential misuse, we also present baseline defense strategies against SpatialJB and evaluate their effectiveness in mitigating such attacks. The code for the attack, baseline defenses, and a demo are available at https://anonymous.4open.science/r/SpatialJailbreak-8E63.

Public large language models (LLMs) are typically safety-aligned during pretraining, yet task-specific fine-tuning required for deployment often erodes this alignment and introduces safety risks. Existing defenses either embed safety recovery into fine-tuning or rely on fine-tuning-derived priors for post-hoc correction, leaving safety recovery tightly coupled with training and incurring high computational overhead and a complex workflow. To address these challenges, we propose \texttt{Q-realign}, a post-hoc defense method based on post-training quantization, guided by an analysis of representational structure. By reframing quantization as a dual-objective procedure for compression and safety, \texttt{Q-realign} decouples safety alignment from fine-tuning and naturally piggybacks into modern deployment pipelines. Experiments across multiple models and datasets demonstrate that our method substantially reduces unsafe behaviors while preserving task performance, with significant reductions in memory usage and GPU hours. Notably, our approach can recover the safety alignment of a fine-tuned 7B LLM on a single RTX 4090 within 40 minutes. Overall, our work provides a practical, turnkey solution for safety-aware deployment.

As Large Language Models (LLMs) are increasingly deployed in healthcare field, it becomes essential to carefully evaluate their medical safety before clinical use. However, existing safety benchmarks remain predominantly English-centric, and test with only single-turn prompts despite multi-turn clinical consultations. To address these gaps, we introduce JMedEthicBench, the first multi-turn conversational benchmark for evaluating medical safety of LLMs for Japanese healthcare. Our benchmark is based on 67 guidelines from the Japan Medical Association and contains over 50,000 adversarial conversations generated using seven automatically discovered jailbreak strategies. Using a dual-LLM scoring protocol, we evaluate 27 models and find that commercial models maintain robust safety while medical-specialized models exhibit increased vulnerability. Furthermore, safety scores decline significantly across conversation turns (median: 9.5 to 5.0, $p < 0.001$). Cross-lingual evaluation on both Japanese and English versions of our benchmark reveals that medical model vulnerabilities persist across languages, indicating inherent alignment limitations rather than language-specific factors. These findings suggest that domain-specific fine-tuning may accidentally weaken safety mechanisms and that multi-turn interactions represent a distinct threat surface requiring dedicated alignment strategies.

Intelligent driving systems are vulnerable to physical adversarial attacks on traffic signs. These attacks can cause misclassification, leading to erroneous driving decisions that compromise road safety. Moreover, within V2X networks, such misinterpretations can propagate, inducing cascading failures that disrupt overall traffic flow and system stability. However, a key limitation of current physical attacks is their lack of stealth. Most methods apply perturbations to central regions of the sign, resulting in visually salient patterns that are easily detectable by human observers, thereby limiting their real-world practicality. This study proposes TESP-Attack, a novel stealth-aware adversarial patch method for traffic sign classification. Based on the observation that human visual attention primarily focuses on the central regions of traffic signs, we employ instance segmentation to generate edge-aligned masks that conform to the shape characteristics of the signs. A U-Net generator is utilized to craft adversarial patches, which are then optimized through color and texture constraints along with frequency domain analysis to achieve seamless integration with the background environment, resulting in highly effective visual concealment. The proposed method demonstrates outstanding attack success rates across traffic sign classification models with varied architectures, achieving over 90% under limited query budgets. It also exhibits strong cross-model transferability and maintains robust real-world performance that remains stable under varying angles and distances.

The emergence of graph foundation models (GFMs), particularly those incorporating language models (LMs), has revolutionized graph learning and demonstrated remarkable performance on text-attributed graphs (TAGs). However, compared to traditional GNNs, these LM-empowered GFMs introduce unique security vulnerabilities during the unsecured prompt tuning phase that remain understudied in current research. Through empirical investigation, we reveal a significant performance degradation in traditional graph backdoor attacks when operating in attribute-inaccessible constrained TAG systems without explicit trigger node attribute optimization. To address this, we propose a novel dual-trigger backdoor attack framework that operates at both text-level and struct-level, enabling effective attacks without explicit optimization of trigger node text attributes through the strategic utilization of a pre-established text pool. Extensive experimental evaluations demonstrate that our attack maintains superior clean accuracy while achieving outstanding attack success rates, including scenarios with highly concealed single-trigger nodes. Our work highlights critical backdoor risks in web-deployed LM-empowered GFMs and contributes to the development of more robust supervision mechanisms for open-source platforms in the era of foundation models.

Large reasoning models (LRMs) with multi-step reasoning capabilities have shown remarkable problem-solving abilities, yet they exhibit concerning safety vulnerabilities that remain poorly understood. In this work, we investigate why safety alignment fails in reasoning models through a mechanistic interpretability lens. Using a linear probing approach to trace refusal intentions across token positions, we discover a striking phenomenon termed as \textbf{refusal cliff}: many poorly-aligned reasoning models correctly identify harmful prompts and maintain strong refusal intentions during their thinking process, but experience a sharp drop in refusal scores at the final tokens before output generation. This suggests that these models are not inherently unsafe; rather, their refusal intentions are systematically suppressed. Through causal intervention analysis, we identify a sparse set of attention heads that negatively contribute to refusal behavior. Ablating just 3\% of these heads can reduce attack success rates below 10\%. Building on these mechanistic insights, we propose \textbf{Cliff-as-a-Judge}, a novel data selection method that identifies training examples exhibiting the largest refusal cliff to efficiently repair reasoning models' safety alignment. This approach achieves comparable safety improvements using only 1.7\% of the vanilla safety training data, demonstrating a less-is-more effect in safety alignment.

Information leakage issues in machine learning-based Web applications have attracted increasing attention. While the risk of data privacy leakage has been rigorously analyzed, the theory of model function leakage, known as Model Extraction Attacks (MEAs), has not been well studied. In this paper, we are the first to understand MEAs theoretically from an attack-agnostic perspective and to propose analytical metrics for evaluating model extraction risks. By using the Neural Tangent Kernel (NTK) theory, we formulate the linearized MEA as a regularized kernel classification problem and then derive the fidelity gap and generalization error bounds of the attack performance. Based on these theoretical analyses, we propose a new theoretical metric called Model Recovery Complexity (MRC), which measures the distance of weight changes between the victim and surrogate models to quantify risk. Additionally, we find that victim model accuracy, which shows a strong positive correlation with model extraction risk, can serve as an empirical metric. By integrating these two metrics, we propose a framework, namely Model Extraction Risk Inspector (MER-Inspector), to compare the extraction risks of models under different model architectures by utilizing relative metric values. We conduct extensive experiments on 16 model architectures and 5 datasets. The experimental results demonstrate that the proposed metrics have a high correlation with model extraction risks, and MER-Inspector can accurately compare the extraction risks of any two models with up to 89.58%.

Safety alignment aims to prevent Large Language Models (LLMs) from responding to harmful queries. To strengthen safety protections, jailbreak methods are developed to simulate malicious attacks and uncover vulnerabilities. In this paper, we introduce HILL (Hiding Intention by Learning from LLMs), a novel jailbreak approach that systematically transforms imperative harmful requests into learning-style questions with only straightforward hypotheticality indicators. Further, we introduce two new metrics to thoroughly evaluate the utility of jailbreak methods. Experiments on the AdvBench dataset across a wide range of models demonstrate HILL's strong effectiveness, generalizability, and harmfulness. It achieves top attack success rates on the majority of models and across malicious categories while maintaining high efficiency with concise prompts. Results of various defense methods show the robustness of HILL, with most defenses having mediocre effects or even increasing the attack success rates. Moreover, the assessment on our constructed safe prompts reveals inherent limitations of LLMs' safety mechanisms and flaws in defense methods. This work exposes significant vulnerabilities of safety measures against learning-style elicitation, highlighting a critical challenge of balancing helpfulness and safety alignments.

The integration of AI with medical images enables the extraction of implicit image-derived biomarkers for a precise health assessment. Recently, retinal age, a biomarker predicted from fundus images, is a proven predictor of systemic disease risks, behavioral patterns, aging trajectory and even mortality. However, the capability to infer such sensitive biometric data raises significant privacy risks, where unauthorized use of fundus images could lead to bioinformation leakage, breaching individual privacy. In response, we formulate a new research problem of biometric privacy associated with medical images and propose RetinaGuard, a novel privacy-enhancing framework that employs a feature-level generative adversarial masking mechanism to obscure retinal age while preserving image visual quality and disease diagnostic utility. The framework further utilizes a novel multiple-to-one knowledge distillation strategy incorporating a retinal foundation model and diverse surrogate age encoders to enable a universal defense against black-box age prediction models. Comprehensive evaluations confirm that RetinaGuard successfully obfuscates retinal age prediction with minimal impact on image quality and pathological feature representation. RetinaGuard is also flexible for extension to other medical image derived biomarkers. RetinaGuard is also flexible for extension to other medical image biomarkers.

Recent advances in vision-language models (VLMs) have significantly enhanced the visual grounding task, which involves locating objects in an image based on natural language queries. Despite these advancements, the security of VLM-based grounding systems has not been thoroughly investigated. This paper reveals a novel and realistic vulnerability: the first multi-target backdoor attack on VLM-based visual grounding. Unlike prior attacks that rely on static triggers or fixed targets, we propose IAG, a method that dynamically generates input-aware, text-guided triggers conditioned on any specified target object description to execute the attack. This is achieved through a text-conditioned UNet that embeds imperceptible target semantic cues into visual inputs while preserving normal grounding performance on benign samples. We further develop a joint training objective that balances language capability with perceptual reconstruction to ensure imperceptibility, effectiveness, and stealth. Extensive experiments on multiple VLMs (e.g., LLaVA, InternVL, Ferret) and benchmarks (RefCOCO, RefCOCO+, RefCOCOg, Flickr30k Entities, and ShowUI) demonstrate that IAG achieves the best ASRs compared with other baselines on almost all settings without compromising clean accuracy, maintaining robustness against existing defenses, and exhibiting transferability across datasets and models. These findings underscore critical security risks in grounding-capable VLMs and highlight the need for further research on trustworthy multimodal understanding.

Prompt-based adversarial attacks have become an effective means to assess the robustness of large language models (LLMs). However, existing approaches often treat prompts as monolithic text, overlooking their structural heterogeneity-different prompt components contribute unequally to adversarial robustness. Prior works like PromptRobust assume prompts are value-neutral, but our analysis reveals that complex, domain-specific prompts with rich structures have components with differing vulnerabilities. To address this gap, we introduce PromptAnatomy, an automated framework that dissects prompts into functional components and generates diverse, interpretable adversarial examples by selectively perturbing each component using our proposed method, ComPerturb. To ensure linguistic plausibility and mitigate distribution shifts, we further incorporate a perplexity (PPL)-based filtering mechanism. As a complementary resource, we annotate four public instruction-tuning datasets using the PromptAnatomy framework, verified through human review. Extensive experiments across these datasets and five advanced LLMs demonstrate that ComPerturb achieves state-of-the-art attack success rates. Ablation studies validate the complementary benefits of prompt dissection and PPL filtering. Our results underscore the importance of prompt structure awareness and controlled perturbation for reliable adversarial robustness evaluation in LLMs. Code and data are available at https://github.com/Yujiaaaaa/PACP.

In recent research, adversarial attacks on person detectors using patches or static 3D model-based texture modifications have struggled with low success rates due to the flexible nature of human movement. Modeling the 3D deformations caused by various actions has been a major challenge. Fortunately, advancements in Neural Radiance Fields (NeRF) for dynamic human modeling offer new possibilities. In this paper, we introduce UV-Attack, a groundbreaking approach that achieves high success rates even with extensive and unseen human actions. We address the challenge above by leveraging dynamic-NeRF-based UV mapping. UV-Attack can generate human images across diverse actions and viewpoints, and even create novel actions by sampling from the SMPL parameter space. While dynamic NeRF models are capable of modeling human bodies, modifying clothing textures is challenging because they are embedded in neural network parameters. To tackle this, UV-Attack generates UV maps instead of RGB images and modifies the texture stacks. This approach enables real-time texture edits and makes the attack more practical. We also propose a novel Expectation over Pose Transformation loss (EoPT) to improve the evasion success rate on unseen poses and views. Our experiments show that UV-Attack achieves a 92.7% attack success rate against the FastRCNN model across varied poses in dynamic video settings, significantly outperforming the state-of-the-art AdvCamou attack, which only had a 28.5% ASR. Moreover, we achieve 49.5% ASR on the latest YOLOv8 detector in black-box settings. This work highlights the potential of dynamic NeRF-based UV mapping for creating more effective adversarial attacks on person detectors, addressing key challenges in modeling human movement and texture modification. The code is available at https://github.com/PolyLiYJ/UV-Attack.