arXiv subjects

Zizhuang Deng

Publications and source records attributed to Zizhuang Deng.

9 recordsLinked to original sources

arXiv2026

EAVer: Long-Form Factuality Verification as an End-to-End Agentic Policy

Long-form factuality verification is commonly implemented as a static decompose-search-verify pipeline, with separately prompted modules processing claims and invoking external search. Treating claims independently makes LLM and search calls scale with claim count and causes repeated searches for overlapping evidence about related claims. We introduce EAVer, an End-to-end Agentic Verifier that learns to control the complete response-level verification workflow as a unified policy. EAVer groups semantically related claims, routes each group to direct verification or targeted search based on confidence, and keeps evidence returned by search in compact in-context memos for cross-claim reuse. To train this policy, we develop a privileged-teacher synthesis pipeline that converts gold claim annotations into executable multi-turn tool-interaction trajectories with live search rather than post-hoc rationales. Structural, label-alignment, tool-use, search-budget, and leakage checks yield 1,447 quality-controlled trajectories. We further construct 794 bidirectional same-trajectory preference pairs that keep claim grouping, search, and evidence fixed, enabling decision-focused Direct Preference Optimization (DPO) over factuality-decision tokens. The results with Qwen3-8B show that EAVer outperforms the strongest search-based baseline on each benchmark by 2.88 Macro-F1 points on VeriFastScore and 4.73 points on the out-of-distribution FaStFact-Bench, while using about 80% fewer searches than the most search-efficient baseline. Moreover, EAVer consistently improves performance across models ranging from 4B to 32B parameters, demonstrating its strong generalizability.

cs.CL↗

arXiv2026

QuantGuard: Learnable Rounding for Repairing Quantization-Conditioned Backdoors in LLMs

Model quantization is a key technique for reducing storage and inference costs in large language model deployment. However, recent studies show that the discretization and rounding errors introduced by quantization can be exploited by adversaries to construct quantization-conditioned backdoor (QCB) attacks. Under such attacks, malicious behavior remains dormant at full precision and activates only after quantization, thereby bypassing conventional security auditing and detection. To address this threat, we propose QuantGuard, a proactive pre-quantization defense that learns safe rounding adjustments through differentiable optimization. Our method introduces differentiable rounding control variables and combines error-guided rounding reversal constraints, output-distribution consistency, and weight-distance regularization to regulate critical rounding behaviors. Crucially, QuantGuard utilizes only a small calibration dataset and does not modify existing quantization algorithms. This design disrupts the alignment between attacker-crafted weight patterns and quantization boundaries, suppressing post-quantization backdoor activation while preserving model functionality and performance. We conduct systematic experiments on six mainstream LLMs (including the LLaMA-3 and Qwen2.5-Coder) using three quantization precisions (INT8, FP4, and NF4) across three representative scenarios: vulnerable code generation, content injection, and over-refusal. The results show that QuantGuard consistently mitigates QCB attacks, reducing the attack success rate to a level comparable to the clean model while largely preserving general capability. With low computational overhead, QuantGuard provides a practical defense for secure quantized LLM deployment.

cs.CR↗

arXiv2026

Beyond Vector Hiding: Breaking and Mitigating Shared-Direction Weight Obfuscation in TEE-Offloaded Large Language Models

Trusted Execution Environment (TEE)-shielded partitioning of Large Language Models (LLMs) accelerates on-device inference by offloading obfuscated linear layers to an untrusted accelerator while retaining only a small correction inside the TEE. However, earlier lightweight obfuscation schemes preserved weight-vector directions and were broken by ArrowMatch. To defend against this attack, ArrowCloak injects scalar multiples of the same hidden direction into all weight vectors, enabling lightweight trusted correction. We show that this reuse leaves a rank-one relation across the complete accelerator-visible matrix. For the released real-valued scheme, we propose SpectralLeak, which estimates and removes the shared component. Across 12 task settings, its surrogates achieve $87.98\%$ mean accuracy versus $89.85\%$ for the victims. In our defense-favorable mod-$Q$ realization of ArrowCloak's published modular security formulation, mod-$Q$ arithmetic suppresses this spectral signal but retains the algebraic rank-one relation modulo $Q$. We therefore propose LatticeLeak, which exploits the resulting hidden lattice. In our BERT-Base and GPT2-Base experiments, it reconstructs every protected fixed-point parameter exactly; across all evaluated architectures, the reconstructed models retain victim-level task accuracy without victim queries, labels, or fine-tuning. These findings identify shared rank-one reuse as the root cause of the leakage exploited by our attacks. Guided by this insight, we design ButterflyCloak, a keyed maximal-rank butterfly mask that replaces the reused direction with distinct mask rows while retaining fast trusted correction...

cs.CR↗

arXiv2026

FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks

Model quantization is essential for the efficient deployment of Large Language Models (LLMs), but introduces a critical vulnerability: Quantization-Conditioned Backdoor (QCB) attacks. In these attacks, malicious behaviors remain dormant in full-precision models and activate only after specific quantization distortions, bypassing standard security audits. To mitigate this, we introduce FlipGuard, a proactive defense framework that selectively perturbs model weights prior to quantization. By breaking the adversary's precise alignment between weight patterns and quantization boundaries, FlipGuard suppresses backdoor activation without requiring access to training data or trigger samples. We further propose the Defense Effectiveness Ratio (DER), a unified metric to jointly evaluate security gains, utility preservation, and computational cost. Extensive experiments across seven LLMs (including StarCoder and LLaMA-family models) and three quantization schemes (INT8, FP4, NF4) demonstrate that FlipGuard effectively neutralizes QCBs across three scenarios, i.e., vulnerable code generation, content injection, and over-refusal, achieving high security with negligible performance degradation.

cs.CR↗

arXiv2026

PlanGuard: Defending Agents against Indirect Prompt Injection via Planning-based Consistency Verification

Large Language Model (LLM) agents are increasingly integrated into critical systems, leveraging external tools to interact with the real world. However, this capability exposes them to Indirect Prompt Injection (IPI), where attackers embed malicious instructions into retrieved content to manipulate the agent into executing unauthorized or unintended actions. Existing defenses predominantly focus on the pre-processing stage, neglecting the monitoring of the model's actual behavior. In this paper, we propose PlanGuard, a training-free defense framework based on the principle of Context Isolation. Unlike prior methods, PlanGuard introduces an isolated Planner that generates a reference set of valid actions derived solely from user instructions. In addition, we design a Hierarchical Verification Mechanism that first enforces strict hard constraints to block unauthorized tool invocations, and subsequently employs an Intent Verifier to validate whether parameter deviations are benign formatting variances or malicious hijacking. Experiments on the InjecAgent benchmark demonstrate that PlanGuard effectively neutralizes these attacks, reducing the Attack Success Rate (ASR) from 72.8% to 0%, while maintaining an acceptable False Positive Rate of 1.49%. Furthermore, our method is model-agnostic and highly compatible.

cs.CR↗

arXiv2025

HLPD: Aligning LLMs to Human Language Preference for Machine-Revised Text Detection

To prevent misinformation and social issues arising from trustworthy-looking content generated by LLMs, it is crucial to develop efficient and reliable methods for identifying the source of texts. Previous approaches have demonstrated exceptional performance in detecting texts fully generated by LLMs. However, these methods struggle when confronting more advanced LLM output or text with adversarial multi-task machine revision, especially in the black-box setting, where the generating model is unknown. To address this challenge, grounded in the hypothesis that human writing possesses distinctive stylistic patterns, we propose Human Language Preference Detection (HLPD). HLPD employs a reward-based alignment process, Human Language Preference Optimization (HLPO), to shift the scoring model's token distribution toward human-like writing, making the model more sensitive to human writing, therefore enhancing the identification of machine-revised text. We test HLPD in an adversarial multi-task evaluation framework that leverages a five-dimensional prompt generator and multiple advanced LLMs to create diverse revision scenarios. When detecting texts revised by GPT-series models, HLPD achieves a 15.11% relative improvement in AUROC over ImBD, surpassing Fast-DetectGPT by 45.56%. When evaluated on texts generated by advanced LLMs, HLPD achieves the highest average AUROC, exceeding ImBD by 5.53% and Fast-DetectGPT by 34.14%. Code will be made available at https://github.com/dfq2021/HLPD.

cs.CL↗

arXiv2025

The Art of Hide and Seek: Making Pickle-Based Model Supply Chain Poisoning Stealthy Again

Pickle deserialization vulnerabilities have persisted throughout Python's history, remaining widely recognized yet unresolved. Due to its ability to transparently save and restore complex objects into byte streams, many AI/ML frameworks continue to adopt pickle as the model serialization protocol despite its inherent risks. As the open-source model ecosystem grows, model-sharing platforms such as Hugging Face have attracted massive participation, significantly amplifying the real-world risks of pickle exploitation and opening new avenues for model supply chain poisoning. Although several state-of-the-art scanners have been developed to detect poisoned models, their incomplete understanding of the poisoning surface leaves the detection logic fragile and allows attackers to bypass them. In this work, we present the first systematic disclosure of the pickle-based model poisoning surface from both model loading and risky function perspectives. Our research demonstrates how pickle-based model poisoning can remain stealthy and highlights critical gaps in current scanning solutions. On the model loading surface, we identify 22 distinct pickle-based model loading paths across five foundational AI/ML frameworks, 19 of which are entirely missed by existing scanners. We further develop a bypass technique named Exception-Oriented Programming (EOP) and discover 9 EOP instances, 7 of which can bypass all scanners. On the risky function surface, we discover 133 exploitable gadgets, achieving almost a 100% bypass rate. Even against the best-performing scanner, these gadgets maintain an 89% bypass rate. By systematically revealing the pickle-based model poisoning surface, we achieve practical and robust bypasses against real-world scanners. We responsibly disclose our findings to corresponding vendors, receiving acknowledgments and a $6000 bug bounty.

cs.CR↗

arXiv2025

Demystifying RCE Vulnerabilities in LLM-Integrated Apps

LLMs show promise in transforming software development, with a growing interest in integrating them into more intelligent apps. Frameworks like LangChain aid LLM-integrated app development, offering code execution utility/APIs for custom actions. However, these capabilities theoretically introduce Remote Code Execution (RCE) vulnerabilities, enabling remote code execution through prompt injections. No prior research systematically investigates these frameworks' RCE vulnerabilities or their impact on applications and exploitation consequences. Therefore, there is a huge research gap in this field. In this study, we propose LLMSmith to detect, validate and exploit the RCE vulnerabilities in LLM-integrated frameworks and apps. To achieve this goal, we develop two novel techniques, including 1) a lightweight static analysis to examine LLM integration mechanisms, and construct call chains to identify RCE vulnerabilities in frameworks; 2) a systematical prompt-based exploitation method to verify and exploit the found vulnerabilities in LLM-integrated apps. This technique involves various strategies to control LLM outputs, trigger RCE vulnerabilities and launch subsequent attacks. Our research has uncovered a total of 20 vulnerabilities in 11 LLM-integrated frameworks, comprising 19 RCE vulnerabilities and 1 arbitrary file read/write vulnerability. Of these, 17 have been confirmed by the framework developers, with 11 vulnerabilities being assigned CVE IDs. For the 51 apps potentially affected by RCE, we successfully executed attacks on 17 apps, 16 of which are vulnerable to RCE and 1 to SQL injection. Furthermore, we conduct a comprehensive analysis of these vulnerabilities and construct practical attacks to demonstrate the hazards in reality. Last, we propose several mitigation measures for both framework and app developers to counteract such attacks.

cs.CR↗

arXiv2022

Understanding Real-world Threats to Deep Learning Models in Android Apps

Famous for its superior performance, deep learning (DL) has been popularly used within many applications, which also at the same time attracts various threats to the models. One primary threat is from adversarial attacks. Researchers have intensively studied this threat for several years and proposed dozens of approaches to create adversarial examples (AEs). But most of the approaches are only evaluated on limited models and datasets (e.g., MNIST, CIFAR-10). Thus, the effectiveness of attacking real-world DL models is not quite clear. In this paper, we perform the first systematic study of adversarial attacks on real-world DNN models and provide a real-world model dataset named RWM. Particularly, we design a suite of approaches to adapt current AE generation algorithms to the diverse real-world DL models, including automatically extracting DL models from Android apps, capturing the inputs and outputs of the DL models in apps, generating AEs and validating them by observing the apps' execution. For black-box DL models, we design a semantic-based approach to build suitable datasets and use them for training substitute models when performing transfer-based attacks. After analyzing 245 DL models collected from 62,583 real-world apps, we have a unique opportunity to understand the gap between real-world DL models and contemporary AE generation algorithms. To our surprise, the current AE generation algorithms can only directly attack 6.53% of the models. Benefiting from our approach, the success rate upgrades to 47.35%.

cs.CR↗