Searcharxiv⌕ Search

arXiv subjects

Zhan Qin

Publications and source records attributed to Zhan Qin.

At least 37 records · Page 2Linked to original sources
arXiv2026

Adaptive and Explicit safe: Triggering Latent Safety Awareness in Large Reasoning Models

While Large Reasoning Models (LRMs) excel at complex tasks, they remain highly vulnerable to sophisticated jailbreaks and direct harmful queries. To address this vulnerability, prior works depend heavily on external manual data annotation for safety alignment. However, we observe that LRMs can inherently identify safety risks when being re-presented with original queries alongside their own reasoning trajectories -- a capability we term Latent Safety Awareness. To leverage this safety awareness, we first employ Supervised Fine-Tuning (SFT) to explicitly induce safe tags to trigger safety analysis and guidance following the initial reasoning content for unsafe queries, while preserving standard responses for general queries to ensure adaptive triggering. Subsequently, we apply Direct Preference Optimization (DPO) to further enhance the correctness and stability of the safety analysis and guidance. Notably, responses required for both training stages are entirely generated by models being optimized. With (Safe Trigger) SFT and DPO, experimental results demonstrate significant safety enhancement. For example, the Attack Success Rate (ASR) of DeepSeek-R1-Distill-Llama-8B, on average, drops 24.65% and 36.72% on harmful and jailbreak benchmarks, respectively. Finally, our Safe Trigger method exerts almost no negative impact on general performance or user experience.

cs.AI↗
arXiv2026

AttriGuard: Defeating Indirect Prompt Injection in LLM Agents via Causal Attribution of Tool Invocations

LLM agents are highly vulnerable to Indirect Prompt Injection (IPI), where adversaries embed malicious directives in untrusted tool outputs to hijack execution. Most existing defenses treat IPI as an input-level semantic discrimination problem, which often fails to generalize to unseen payloads. We propose a new paradigm, action-level causal attribution, which secures agents by asking why a particular tool call is produced. The central goal is to distinguish tool calls supported by the user's intent from those causally driven by untrusted observations. We instantiate this paradigm with AttriGuard, a runtime defense based on parallel counterfactual tests. For each proposed tool call, AttriGuard verifies its necessity by re-executing the agent under a control-attenuated view of external observations. Technically, AttriGuard combines teacher-forced shadow replay to prevent attribution confounding, hierarchical control attenuation to suppress diverse control channels while preserving task-relevant information, and a fuzzy survival criterion that is robust to LLM stochasticity. Across four LLMs and two agent benchmarks, AttriGuard achieves 0% ASR under static attacks with negligible utility loss and moderate overhead. Importantly, it remains resilient under adaptive optimization-based attacks in settings where leading defenses degrade significantly.

cs.CR↗
arXiv2026

FIT-Print: Towards False-claim-resistant Model Ownership Verification via Targeted Fingerprint

Model fingerprinting has emerged as a crucial mechanism for safeguarding the intellectual property of open-source models, offering a non-intrusive approach that requires no modifications to the protected model. However, our analysis reveals that existing fingerprinting techniques are fundamentally vulnerable to false claim attacks, wherein adversaries can fraudulently assert ownership over independent third-party models. We demonstrate that this vulnerability stems from the untargeted nature of current methods, which evaluate model similarity based on arbitrary sample outputs rather than alignment with a specific, predefined reference. To mitigate this vulnerability, we introduce FIT-Print, a targeted fingerprinting paradigm that actively counters false claim attacks. Specifically, FIT-Print leverages optimization to transform the fingerprint into a verifiable, targeted signature. Building upon this foundation, we propose two black-box fingerprinting methods, the bit-wise FIT-ModelDiff and the list-wise FIT-LIME, which utilize output distances and feature attributions as robust model signatures, respectively. Extensive evaluations across benchmark models and datasets show that our framework perfectly neutralizes false claim attacks (100% defense success rate) and eliminates false alarms on independent models (0.0%), all while maintaining a 100% ownership verification rate against diverse model reuse techniques.

cs.CR↗
arXiv2026

PromptCOS: Towards Content-only System Prompt Copyright Auditing for LLMs

System prompts are critical for shaping the behavior and output quality of large language model (LLM)-based applications, driving substantial investment in optimizing high-quality prompts beyond traditional handcrafted designs. However, as system prompts become valuable intellectual property, they are increasingly vulnerable to prompt theft and unauthorized use, highlighting the urgent need for effective copyright auditing, especially watermarking. Existing methods rely on verifying subtle logit distribution shifts triggered by a query. We observe that this logit-dependent verification framework is impractical in real-world content-only settings, primarily because (1) random sampling makes content-level generation unstable for verification, and (2) stronger instructions needed for content-level signals compromise prompt fidelity. To overcome these challenges, we propose PromptCOS, the first content-only system prompt copyright auditing method based on content-level output similarity. PromptCOS achieves watermark stability by designing a cyclic output signal as the conditional instruction's target. It preserves prompt fidelity by injecting a small set of auxiliary tokens to encode the watermark, leaving the main prompt untouched. Furthermore, to ensure robustness against malicious removal, we optimize cover tokens, i.e., critical tokens in the original prompt, to ensure that removing auxiliary tokens causes severe performance degradation. Experimental results show that promptCOS achieves high effectiveness (99.3% average watermark similarity), strong distinctiveness (60.8% higher than the best baseline), high fidelity (accuracy degradation no greater than 0.6%), robustness (resilience against four potential attack categories), and high computational efficiency (up to 98.1% cost saving).

cs.CR↗
arXiv2026

EVA: Editing for Versatile Alignment against Jailbreaks

Large Language Models (LLMs) and Vision Language Models (VLMs) have demonstrated impressive capabilities but remain vulnerable to jailbreaking attacks, where adversaries exploit textual or visual triggers to bypass safety guardrails. Recent defenses typically rely on safety fine-tuning or external filters to reduce the model's likelihood of producing harmful content. While effective to some extent, these methods often incur significant computational overheads and suffer from the safety utility trade-off, degrading the model's performance on benign tasks. To address these challenges, we propose EVA (Editing for Versatile Alignment against Jailbreaks), a novel framework that pioneers the application of direct model editing for safety alignment. EVA reframes safety alignment as a precise knowledge correction task. Instead of retraining massive parameters, EVA identifies and surgically edits specific neurons responsible for the model's susceptibility to harmful instructions, while leaving the vast majority of the model unchanged. By localizing the updates, EVA effectively neutralizes harmful behaviors without compromising the model's general reasoning capabilities. Extensive experiments demonstrate that EVA outperforms baselines in mitigating jailbreaks across both LLMs and VLMs, offering a precise and efficient solution for post-deployment safety alignment.

cs.CR↗
arXiv2026

Defense against Poisoning Attacks under Shuffle-DP

Differential Privacy (DP) has become the gold standard for protecting individual privacy in data analytics, and the shuffle-DP model has attracted significant attention from both academia and industry due to its favorable balance between privacy and utility. However, existing shuffle-DP protocols rely on a strong assumption: all users behave honestly. In real-world scenarios, adversarial users can exploit this vulnerability through poisoning attacks, compromising both privacy guarantees and the utility of analytical results. While defending against poisoning attacks in the shuffle-DP model has recently gained interest, existing solutions are limited to frequency estimation tasks. To address this issue, we propose the first general defense framework for all union-preserving queries, capable of transforming any shuffle-DP protocol into a version resilient to poisoning attacks. Beyond robust defense against poisoning attacks, our framework achieves high utility of analytical results. Compared to the original shuffle-DP protocol, it retains asymptotically equivalent error in attack-free settings and incurs only a polylogarithmic increase in error when a constant number of attackers are present. We demonstrate the generality of our framework on several common queries, including summation, frequency estimation, and range counting. Experimental results confirm that our approach effectively defends against poisoning attacks while maintaining strong utility and communication efficiency.

cs.CR↗
arXiv2026

Channel-Level Semantic Perturbations: Unlearnable Examples for Diverse Training Paradigms

The unauthorized use of personal data in model training has emerged as a growing privacy threat. Unlearnable examples (UEs) address this issue by embedding imperceptible perturbations into benign examples to obstruct feature learning. However, existing studies mainly evaluate UEs under from-scratch training settings, leaving their behavior under the widely adopted pretraining-finetuning (PF) paradigm largely unexplored. In this work, we provide the first systematic investigation of unlearnable examples across diverse training paradigms. Our analysis reveals that loading and freezing pretrained weights significantly weakens the effectiveness of existing UEs methods. We further explain these findings through semantic filtering: while UEs tend to induce models to overfit non-semantic noise, thereby weakening their semantic extraction capabilities, under the PF paradigm, frozen shallow layers preserve data semantics, effectively filtering out distracting information like unlearnable noise. Guided by these insights, we propose a hierarchical deception strategy, Shallow Semantic Camouflage (SSC), that confines the generation process to a semantically valid subspace, aiming to bypass the semantic suppression introduced by pretrained weights. Extensive experiments demonstrate that our method consistently preserves data unlearnability even under challenging training paradigms, such as shallow-layer freezing and semantic-focused pretraining (SF-Pretrain), bridging the critical gap in pretrain-based unlearnable learning.

cs.LG↗
arXiv2026

Towards Identification and Intervention of Safety-Critical Parameters in Large Language Models

Ensuring Large Language Model (LLM) safety is crucial, yet the lack of a clear understanding about safety mechanisms hinders the development of precise and reliable methodologies for safety intervention across diverse tasks. To better understand and control LLM safety, we propose the Expected Safety Impact (ESI) framework for quantifying how different parameters affect LLM safety. Based on ESI, we reveal distinct safety-critical patterns across different LLM architectures: In dense LLMs, many safety-critical parameters are located in value matrices (V) and MLPs in middle layers, whereas in Mixture-of-Experts (MoE) models, they shift to the late-layer MLPs. Leveraging ESI, we further introduce two targeted intervention paradigms for safety enhancement and preservation, i.e., Safety Enhancement Tuning (SET) and Safety Preserving Adaptation (SPA). SET can align unsafe LLMs by updating only a few safety-critical parameters, effectively enhancing safety while preserving original performance. SPA safeguards well-aligned LLMs during capability-oriented intervention (e.g., instruction tuning) by preventing disruption of safety-critical weights, allowing the LLM to acquire new abilities and maintain safety capabilities. Extensive evaluations on different LLMs demonstrate that SET can reduce the attack success rates of unaligned LLMs by over 50% with only a 100-iteration update on 1% of model weights. SPA can limit the safety degradation of aligned LLMs within 1% after a 1,000-iteration instruction fine-tuning on different tasks. Our code is available at: https://github.com/ZJU-LLM-Safety/SafeWeights-ACL.

cs.CR↗
arXiv2026

JANUS: A Lightweight Framework for Jailbreaking Text-to-Image Models via Distribution Optimization

Text-to-image (T2I) models such as Stable Diffusion and DALLE remain susceptible to generating harmful or Not-Safe-For-Work (NSFW) content under jailbreak attacks despite deployed safety filters. Existing jailbreak attacks either rely on proxy-loss optimization instead of the true end-to-end objective, or depend on large-scale and costly RL-trained generators. Motivated by these limitations, we propose JANUS , a lightweight framework that formulates jailbreak as optimizing a structured prompt distribution under a black-box, end-to-end reward from the T2I system and its safety filters. JANUS replaces a high-capacity generator with a low-dimensional mixing policy over two semantically anchored prompt distributions, enabling efficient exploration while preserving the target semantics. On modern T2I models, we outperform state-of-the-art jailbreak methods, improving ASR-8 from 25.30% to 43.15% on Stable Diffusion 3.5 Large Turbo with consistently higher CLIP and NSFW scores. JANUS succeeds across both open-source and commercial models. These findings expose structural weaknesses in current T2I safety pipelines and motivate stronger, distribution-aware defenses. Warning: This paper contains model outputs that may be offensive.

cs.CV↗
arXiv2026

HarmMetric Eval: Benchmarking Metrics and Judges for LLM Harmfulness Assessment

The potential of large language models (LLMs) to generate harmful content poses a significant safety risk for data management, as LLMs are increasingly being used as engines for data generation. To assess this risk, numerous harmfulness evaluation metrics and judges have been proposed. However, due to differences in their formats and scales, these metrics may yield inconsistent evaluation results on LLM-generated harmful data, undermining their credibility in practice. To address this gap, we present HarmMetric Eval, a systematic benchmark for assessing the quality of harmfulness metrics and judges with varying formats and scales. HarmMetric Eval includes a high-quality dataset comprising representative harmful prompts paired with harmful and non-harmful LLM outputs across multiple fine-grained categories, along with a unified scoring mechanism to reward the metrics for correctly ranking harmful outputs over non-harmful ones. Extensive experiments on HarmMetric Eval yield a surprising finding: conventional reference-based metrics such as ROUGE and METEOR can outperform LLM-based judges in fine-grained harmfulness evaluation, challenging prevailing assumptions about LLMs' superiority in this domain. To reveal the reasons behind this finding, we provide a fine-grained analysis to explain the limitations of LLM-based judges on rating irrelevant or useless LLM outputs. Motivated by these insights, we design an improved harmfulness judge that explicitly incorporates fine-grained harmfulness criteria in its prompt template and leverages reference-based metrics for lightweight fine-tuning of its base LLM. The resulting judge achieves state-of-the-art evaluation effectiveness on HarmMetric Eval.

cs.CL↗
arXiv2026

Shadow in the Cache: Unveiling and Mitigating Privacy Risks of KV-cache in LLM Inference

The Key-Value (KV) cache, which stores intermediate attention computations (Key and Value pairs) to avoid redundant calculations, is a fundamental mechanism for accelerating Large Language Model (LLM) inference. However, this efficiency optimization introduces significant yet underexplored privacy risks. This paper provides the first comprehensive analysis of these vulnerabilities, demonstrating that an attacker can reconstruct sensitive user inputs directly from the KV-cache. We design and implement three distinct attack vectors: a direct Inversion Attack, a more broadly applicable and potent Collision Attack, and a semantic-based Injection Attack. These methods demonstrate the practicality and severity of KV-cache privacy leakage issues. To mitigate this, we propose KV-Cloak, a novel, lightweight, and efficient defense mechanism. KV-Cloak uses a reversible matrix-based obfuscation scheme, combined with operator fusion, to secure the KV-cache. Our extensive experiments show that KV-Cloak effectively thwarts all proposed attacks, reducing reconstruction quality to random noise. Crucially, it achieves this robust security with virtually no degradation in model accuracy and minimal performance overhead, offering a practical solution for trustworthy LLM deployment.

cs.CR↗
arXiv2026

SmartGuard: Leveraging Large Language Models for Network Attack Detection through Audit Log Analysis and Summarization

End-point monitoring solutions are widely deployed in today's enterprise environments to support advanced attack detection and investigation. These monitors continuously record system-level activities as audit logs and provide deep visibility into security events. Unfortunately, existing methods of semantic analysis based on audit logs have low granularity, only reaching the system call level, making it difficult to effectively classify highly covert behaviors. Additionally, existing works mainly match audit log streams with rule knowledge bases describing behaviors, which heavily rely on expertise and lack the ability to detect unknown attacks and provide interpretive descriptions. In this paper, we propose SmartGuard, an automated method that combines abstracted behaviors from audit event semantics with large language models. SmartGuard extracts specific behaviors (function level) from incoming system logs and constructs a knowledge graph, divides events by threads, and combines event summaries with graph embeddings to achieve information diagnosis and provide explanatory narratives through large language models. Our evaluation shows that SmartGuard achieves an average F1 score of 96\% in assessing malicious behaviors and demonstrates good scalability across multiple models and unknown attacks. It also possesses excellent fine-tuning capabilities, allowing experts to assist in timely system updates.

cs.CR↗
arXiv2026

DualMind: Towards Understanding Cognitive-Affective Cascades in Public Opinion Dissemination via Multi-Agent Simulation

Forecasting public opinion during PR crises is challenging, as existing frameworks often overlook the interaction between transient affective responses and persistent cognitive beliefs. To address this, we propose DualMind, an LLM-driven multi-agent platform designed to model this dual-component interplay. We evaluate the system on 15 real-world crises occurring post-August 2024 using social media data as ground truth. Empirical results demonstrate that DualMind faithfully reconstructs opinion trajectories, significantly outperforming state-of-the-art baselines. This work offers a high-fidelity tool for proactive crisis management. Code is available at https://github.com/EonHao/DualMind.

cs.SI↗
arXiv2026

Reading Between the Lines: Towards Reliable Black-box LLM Fingerprinting via Zeroth-order Gradient Estimation

The substantial investment required to develop Large Language Models (LLMs) makes them valuable intellectual property, raising significant concerns about copyright protection. LLM fingerprinting has emerged as a key technique to address this, which aims to verify a model's origin by extracting an intrinsic, unique signature (a "fingerprint") and comparing it to that of a source model to identify illicit copies. However, existing black-box fingerprinting methods often fail to generate distinctive LLM fingerprints. This ineffectiveness arises because black-box methods typically rely on model outputs, which lose critical information about the model's unique parameters due to the usage of non-linear functions. To address this, we first leverage Fisher Information Theory to formally demonstrate that the gradient of the model's input is a more informative feature for fingerprinting than the output. Based on this insight, we propose ZeroPrint, a novel method that approximates these information-rich gradients in a black-box setting using zeroth-order estimation. ZeroPrint overcomes the challenge of applying this to discrete text by simulating input perturbations via semantic-preserving word substitutions. This operation allows ZeroPrint to estimate the model's Jacobian matrix as a unique fingerprint. Experiments on the standard benchmark show ZeroPrint achieves a state-of-the-art effectiveness and robustness, significantly outperforming existing black-box methods.

cs.CR↗
arXiv2025

DATABench: Evaluating Dataset Auditing in Deep Learning from an Adversarial Perspective

The widespread application of Deep Learning across diverse domains hinges critically on the quality and composition of training datasets. However, the common lack of disclosure regarding their usage raises significant privacy and copyright concerns. Dataset auditing techniques, which aim to determine if a specific dataset was used to train a given suspicious model, provide promising solutions to addressing these transparency gaps. While prior work has developed various auditing methods, their resilience against dedicated adversarial attacks remains largely unexplored. To bridge the gap, this paper initiates a comprehensive study evaluating dataset auditing from an adversarial perspective. We start with introducing a novel taxonomy, classifying existing methods based on their reliance on internal features (IF) (inherent to the data) versus external features (EF) (artificially introduced for auditing). Subsequently, we formulate two primary attack types: evasion attacks, designed to conceal the use of a dataset, and forgery attacks, intending to falsely implicate an unused dataset. Building on the understanding of existing methods and attack objectives, we further propose systematic attack strategies: decoupling, removal, and detection for evasion; adversarial example-based methods for forgery. These formulations and strategies lead to our new benchmark, DATABench, comprising 17 evasion attacks, 5 forgery attacks, and 9 representative auditing methods. Extensive evaluations using DATABench reveal that none of the evaluated auditing methods are sufficiently robust or distinctive under adversarial settings. These findings underscore the urgent need for developing a more secure and reliable dataset auditing method capable of withstanding sophisticated adversarial manipulation. Code is available in https://github.com/shaoshuo-ss/DATABench.

cs.CR↗
arXiv2025

DREAM: Scalable Red Teaming for Text-to-Image Generative Systems via Distribution Modeling

Despite the integration of safety alignment and external filters, text-to-image (T2I) generative systems are still susceptible to producing harmful content, such as sexual or violent imagery. This raises serious concerns about unintended exposure and potential misuse. Red teaming, which aims to proactively identify diverse prompts that can elicit unsafe outputs from the T2I system, is increasingly recognized as an essential method for assessing and improving safety before real-world deployment. However, existing automated red teaming approaches often treat prompt discovery as an isolated, prompt-level optimization task, which limits their scalability, diversity, and overall effectiveness. To bridge this gap, in this paper, we propose DREAM, a scalable red teaming framework to automatically uncover diverse problematic prompts from a given T2I system. Unlike prior work that optimizes prompts individually, DREAM directly models the probabilistic distribution of the target system's problematic prompts, which enables explicit optimization over both effectiveness and diversity, and allows efficient large-scale sampling after training. To achieve this without direct access to representative training samples, we draw inspiration from energy-based models and reformulate the objective into a simple and tractable form. We further introduce GC-SPSA, an efficient optimization algorithm that provides stable gradient estimates through the long and potentially non-differentiable T2I pipeline. During inference, we also propose a diversity-aware sampling strategy to enhance prompt variety. The effectiveness of DREAM is validated through extensive experiments, demonstrating state-of-the-art performance across a wide range of T2I models and safety filters in terms of both prompt success rate and diversity. Our code is available at https://github.com/AntigoneRandy/DREAM

cs.CR↗
arXiv2025

Towards Resilient Safety-driven Unlearning for Diffusion Models against Downstream Fine-tuning

Text-to-image (T2I) diffusion models have achieved impressive image generation quality and are increasingly fine-tuned for personalized applications. However, these models often inherit unsafe behaviors from toxic pretraining data, raising growing safety concerns. While recent safety-driven unlearning methods have made promising progress in suppressing model toxicity, they are found to be fragile to downstream fine-tuning, as we reveal that state-of-the-art methods largely fail to retain their effectiveness even when fine-tuned on entirely benign datasets. To mitigate this problem, in this paper, we propose ResAlign, a safety-driven unlearning framework with enhanced resilience against downstream fine-tuning. By modeling downstream fine-tuning as an implicit optimization problem with a Moreau envelope-based reformulation, ResAlign enables efficient gradient estimation to minimize the recovery of harmful behaviors. Additionally, a meta-learning strategy is proposed to simulate a diverse distribution of fine-tuning scenarios to improve generalization. Extensive experiments across a wide range of datasets, fine-tuning methods, and configurations demonstrate that ResAlign consistently outperforms prior unlearning approaches in retaining safety, while effectively preserving benign generation capability. Our code and pretrained models are publicly available at https://github.com/AntigoneRandy/ResAlign.

cs.LG↗
arXiv2025

Eguard: Defending LLM Embeddings Against Inversion Attacks via Text Mutual Information Optimization

Embeddings have become a cornerstone in the functionality of large language models (LLMs) due to their ability to transform text data into rich, dense numerical representations that capture semantic and syntactic properties. These embedding vector databases serve as the long-term memory of LLMs, enabling efficient handling of a wide range of natural language processing tasks. However, the surge in popularity of embedding vector databases in LLMs has been accompanied by significant concerns about privacy leakage. Embedding vector databases are particularly vulnerable to embedding inversion attacks, where adversaries can exploit the embeddings to reverse-engineer and extract sensitive information from the original text data. Existing defense mechanisms have shown limitations, often struggling to balance security with the performance of downstream tasks. To address these challenges, we introduce Eguard, a novel defense mechanism designed to mitigate embedding inversion attacks. Eguard employs a transformer-based projection network and text mutual information optimization to safeguard embeddings while preserving the utility of LLMs. Our approach significantly reduces privacy risks, protecting over 95% of tokens from inversion while maintaining high performance across downstream tasks consistent with original embeddings.

cs.CR↗