arXiv subjects

Guozhu Meng

Publications and source records attributed to Guozhu Meng.

At least 19 recordsLinked to original sources

arXiv2026

AdaptPrint: Response-Adaptive Fingerprinting of Black-Box LLM Services

Black-box LLM services have emerged as a practical deployment paradigm. Nevertheless, their opacity also hinders the systematic assessment of security risks and complicates copyright auditing for model owners. Black-box LLM fingerprinting, which identifies the underlying LLM identity through query-response interactions, offers a promising way to bridge this gap. Existing approaches typically collect responses from target LLM services using a fixed set of queries and perform poorly in the presence of realistic and complex configurations (e.g., system prompt and sampling settings). To overcome these limitations, we propose AdaptPrint, a response-adaptive fingerprinting method for revealing hidden LLM identities in black-box LLM services. AdaptPrint integrates three progressive response consistency probing strategies: Direct Probing, Continuation Probing, and Follow-up Probing. AdaptPrint determines the final LLM identity by performing similarity matching among candidate LLMs. Experimental results show that AdaptPrint significantly outperforms state-of-the-art methods among 27 candidate models, achieving Top-1, Top-3, and Top-5 accuracies of 80.6%, 90.3%, and 92.1%. AdaptPrint also demonstrates strong robustness across different defense strategies and decoding parameters.

cs.CR↗

arXiv2026

DREA: Decoupled Reasoning and Exploration Agents for Repository-Level Vulnerability Detection

Large language models (LLMs) are increasingly applied to vulnerability detection due to their strong code comprehension capabilities, but most existing approaches rely on isolated functions or context extracted by fixed program-analysis rules. These methods cannot adaptively explore repository-level dependencies to gather sufficient context when vulnerabilities span multiple functions or files, compromising detection reliability. We present DREA (Decoupled Reasoning and Exploration Agents), a hypothesis-driven framework for repository-level vulnerability detection. DREA decouples reasoning from exploration through two collaborating agents: a planning agent backed by an advanced LLM that forms vulnerability hypotheses and directs the investigation, and an explorer agent powered by a lightweight model that retrieves repository-level context on demand. Goal-directed context acquisition is the primary source of detection improvement in this design, while offloading token-heavy exploration to the local model keeps inference economically tractable. To support evaluation, we construct RepoPairBench, a repository-grounded benchmark of validated Python vulnerability-fix pairs from real-world projects. Beyond binary detection accuracy, we introduce a reasoning correctness evaluation to assess whether a model's rationale matches the documented vulnerability mechanism. Across three LLMs, DREA improves Pair-Correctness from 19-26% to 30-42% while offloading over 93% of tokens to the explorer, reducing estimated billable API cost by a factor of 16-48. Reasoning correctness analysis further reveals that 26-55% of true positives, for both DREA and the function-only baseline, are correct predictions supported by flawed rationales, identifying security reasoning quality as a shared bottleneck for current LLMs.

cs.CR↗

arXiv2025

NeuroDeX: Unlocking Diverse Support in Decompiling Deep Neural Network Executables

On-device deep learning models have extensive real world demands. Deep learning compilers efficiently compile models into executables for deployment on edge devices, but these executables may face the threat of reverse engineering. Previous studies have attempted to decompile DNN executables, but they face challenges in handling compilation optimizations and analyzing quantized compiled models. In this paper, we present NeuroDeX to unlock diverse support in decompiling DNN executables. NeuroDeX leverages the semantic understanding capabilities of LLMs along with dynamic analysis to accurately and efficiently perform operator type recognition, operator attribute recovery and model reconstruction. NeuroDeX can recover DNN executables into high-level models towards compilation optimizations, different architectures and quantized compiled models. We conduct experiments on 96 DNN executables across 12 common DNN models. Extensive experimental results demonstrate that NeuroDeX can decompile non-quantized executables into nearly identical high-level models. NeuroDeX can recover functionally similar high-level models for quantized executables, achieving an average top-1 accuracy of 72%. NeuroDeX offers a more comprehensive and effective solution compared to previous DNN executables decompilers.

cs.LG↗

arXiv2025

When Code Crosses Borders: A Security-Centric Study of LLM-based Code Translation

Code translation is crucial for cross-language codebase migration, and large language models (LLMs) have emerged as a promising technique to automate this process. However, the security implications of using LLMs for code translation remain largely unexplored, as existing evaluations primarily focus on syntactic and functional correctness. To bridge this gap, we conduct a security-centric empirical study to investigate the risks of vulnerabilities being introduced or preserved during LLM-based translation. Our study involves a rigorous evaluation of five state-of-the-art LLMs on a curated dataset of 720 security-related code samples across four programming languages (Java, PHP, C, C++) and nine Common Weakness Enumeration (CWE) categories. The results reveal significant security degradation, with 28.6\% to 45\% of translations introducing new vulnerabilities. Web-related flaws, particularly in input validation, proved most challenging for LLMs. Furthermore, we identify and categorize the root causes of these vulnerable translations into a taxonomy of five major error types. Based on our findings, we develop and evaluate a Retrieval-Augmented Generation (RAG)-based mitigation strategy, which successfully reduces the vulnerability introduction rate by 32.8\%. Our study provides the first large-scale evidence of serious security risks in LLM-based code translation and demonstrates the potential of knowledge-enhanced prompting to mitigate them.

cs.CR↗

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

Measuring and Explaining the Effects of Android App Transformations in Online Malware Detection

It is well known that antivirus engines are vulnerable to evasion techniques (e.g., obfuscation) that transform malware into its variants. However, it cannot be necessarily attributed to the effectiveness of these evasions, and the limits of engines may also make this unsatisfactory result. In this study, we propose a data-driven approach to measure the effect of app transformations to malware detection, and further explain why the detection result is produced by these engines. First, we develop an interaction model for antivirus engines, illustrating how they respond with different detection results in terms of varying inputs. Six app transformation techniques are implemented in order to generate a large number of Android apps with traceable changes. Then we undertake a one-month tracking of app detection results from multiple antivirus engines, through which we obtain over 971K detection reports from VirusTotal for 179K apps in total. Last, we conduct a comprehensive analysis of antivirus engines based on these reports from the perspectives of signature-based, static analysis-based, and dynamic analysis-based detection techniques. The results, together with 7 highlighted findings, identify a number of sealed working mechanisms occurring inside antivirus engines and what are the indicators of compromise in apps during malware detection.

cs.CR↗

arXiv2025

Bleeding Pathways: Vanishing Discriminability in LLM Hidden States Fuels Jailbreak Attacks

LLMs remain vulnerable to jailbreak attacks that exploit adversarial prompts to circumvent safety measures. Current safety fine-tuning approaches face two critical limitations. First, they often fail to strike a balance between security and utility, where stronger safety measures tend to over-reject harmless user requests. Second, they frequently miss malicious intent concealed within seemingly benign tasks, leaving models exposed to exploitation. Our work identifies a fundamental cause of these issues: during response generation, an LLM's capacity to differentiate harmful from safe outputs deteriorates. Experimental evidence confirms this, revealing that the separability between hidden states for safe and harmful responses diminishes as generation progresses. This weakening discrimination forces models to make compliance judgments earlier in the generation process, restricting their ability to recognize developing harmful intent and contributing to both aforementioned failures. To mitigate this vulnerability, we introduce DEEPALIGN - an inherent defense framework that enhances the safety of LLMs. By applying contrastive hidden-state steering at the midpoint of response generation, DEEPALIGN amplifies the separation between harmful and benign hidden states, enabling continuous intrinsic toxicity detection and intervention throughout the generation process. Across diverse LLMs spanning varying architectures and scales, it reduced attack success rates of nine distinct jailbreak attacks to near-zero or minimal. Crucially, it preserved model capability while reducing over-refusal. Models equipped with DEEPALIGN exhibited up to 3.5% lower error rates in rejecting challenging benign queries and maintained standard task performance with less than 1% decline. This marks a substantial advance in the safety-utility Pareto frontier.

cs.CR↗

arXiv2025

SAP-DIFF: Semantic Adversarial Patch Generation for Black-Box Face Recognition Models via Diffusion Models

Given the need to evaluate the robustness of face recognition (FR) models, many efforts have focused on adversarial patch attacks that mislead FR models by introducing localized perturbations. Impersonation attacks are a significant threat because adversarial perturbations allow attackers to disguise themselves as legitimate users. This can lead to severe consequences, including data breaches, system damage, and misuse of resources. However, research on such attacks in FR remains limited. Existing adversarial patch generation methods exhibit limited efficacy in impersonation attacks due to (1) the need for high attacker capabilities, (2) low attack success rates, and (3) excessive query requirements. To address these challenges, we propose a novel method SAP-DIFF that leverages diffusion models to generate adversarial patches via semantic perturbations in the latent space rather than direct pixel manipulation. We introduce an attention disruption mechanism to generate features unrelated to the original face, facilitating the creation of adversarial samples and a directional loss function to guide perturbations toward the target identity feature space, thereby enhancing attack effectiveness and efficiency. Extensive experiments on popular FR models and datasets demonstrate that our method outperforms state-of-the-art approaches, achieving an average attack success rate improvement of 45.66% (all exceeding 40%), and a reduction in the number of queries by about 40% compared to the SOTA approach

cs.CV↗

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↗

arXiv2025

Dormant: Defending against Pose-driven Human Image Animation

Pose-driven human image animation has achieved tremendous progress, enabling the generation of vivid and realistic human videos from just one single photo. However, it conversely exacerbates the risk of image misuse, as attackers may use one available image to create videos involving politics, violence, and other illegal content. To counter this threat, we propose Dormant, a novel protection approach tailored to defend against pose-driven human image animation techniques. Dormant applies protective perturbation to one human image, preserving the visual similarity to the original but resulting in poor-quality video generation. The protective perturbation is optimized to induce misextraction of appearance features from the image and create incoherence among the generated video frames. Our extensive evaluation across 8 animation methods and 4 datasets demonstrates the superiority of Dormant over 6 baseline protection methods, leading to misaligned identities, visual distortions, noticeable artifacts, and inconsistent frames in the generated videos. Moreover, Dormant shows effectiveness on 6 real-world commercial services, even with fully black-box access.

cs.CR↗

arXiv2025

Decictor: Towards Evaluating the Robustness of Decision-Making in Autonomous Driving Systems

Autonomous Driving System (ADS) testing is crucial in ADS development, with the current primary focus being on safety. However, the evaluation of non-safety-critical performance, particularly the ADS's ability to make optimal decisions and produce optimal paths for autonomous vehicles (AVs), is also vital to ensure the intelligence and reduce risks of AVs. Currently, there is little work dedicated to assessing the robustness of ADSs' path-planning decisions (PPDs), i.e., whether an ADS can maintain the optimal PPD after an insignificant change in the environment. The key challenges include the lack of clear oracles for assessing PPD optimality and the difficulty in searching for scenarios that lead to non-optimal PPDs. To fill this gap, in this paper, we focus on evaluating the robustness of ADSs' PPDs and propose the first method, Decictor, for generating non-optimal decision scenarios (NoDSs), where the ADS does not plan optimal paths for AVs. Decictor comprises three main components: Non-invasive Mutation, Consistency Check, and Feedback. To overcome the oracle challenge, Non-invasive Mutation is devised to implement conservative modifications, ensuring the preservation of the original optimal path in the mutated scenarios. Subsequently, the Consistency Check is applied to determine the presence of non-optimal PPDs by comparing the driving paths in the original and mutated scenarios. To deal with the challenge of large environment space, we design Feedback metrics that integrate spatial and temporal dimensions of the AV's movement. These metrics are crucial for effectively steering the generation of NoDSs. We evaluate Decictor on Baidu Apollo, an open-source and production-grade ADS. The experimental results validate the effectiveness of Decictor in detecting non-optimal PPDs of ADSs.

cs.AI↗

arXiv2024

Deep Learning-based Software Engineering: Progress, Challenges, and Opportunities

Researchers have recently achieved significant advances in deep learning techniques, which in turn has substantially advanced other research disciplines, such as natural language processing, image processing, speech recognition, and software engineering. Various deep learning techniques have been successfully employed to facilitate software engineering tasks, including code generation, software refactoring, and fault localization. Many papers have also been presented in top conferences and journals, demonstrating the applications of deep learning techniques in resolving various software engineering tasks. However, although several surveys have provided overall pictures of the application of deep learning techniques in software engineering, they focus more on learning techniques, that is, what kind of deep learning techniques are employed and how deep models are trained or fine-tuned for software engineering tasks. We still lack surveys explaining the advances of subareas in software engineering driven by deep learning techniques, as well as challenges and opportunities in each subarea. To this end, in this paper, we present the first task-oriented survey on deep learning-based software engineering. It covers twelve major software engineering subareas significantly impacted by deep learning techniques. Such subareas spread out the through the whole lifecycle of software development and maintenance, including requirements engineering, software development, testing, maintenance, and developer collaboration. As we believe that deep learning may provide an opportunity to revolutionize the whole discipline of software engineering, providing one survey covering as many subareas as possible in software engineering can help future research push forward the frontier of deep learning-based software engineering more systematically.

cs.SE↗

arXiv2024

A Survey on Physical Adversarial Attacks against Face Recognition Systems

As Face Recognition (FR) technology becomes increasingly prevalent in finance, the military, public safety, and everyday life, security concerns have grown substantially. Physical adversarial attacks targeting FR systems in real-world settings have attracted considerable research interest due to their practicality and the severe threats they pose. However, a systematic overview focused on physical adversarial attacks against FR systems is still lacking, hindering an in-depth exploration of the challenges and future directions in this field. In this paper, we bridge this gap by comprehensively collecting and analyzing physical adversarial attack methods targeting FR systems. Specifically, we first investigate the key challenges of physical attacks on FR systems. We then categorize existing physical attacks into three categories based on the physical medium used and summarize how the research in each category has evolved to address these challenges. Furthermore, we review current defense strategies and discuss potential future research directions. Our goal is to provide a fresh, comprehensive, and deep understanding of physical adversarial attacks against FR systems, thereby inspiring relevant research in this area.

cs.CR↗

arXiv2024

Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction

In recent years, large language models (LLMs) have demonstrated notable success across various tasks, but the trustworthiness of LLMs is still an open problem. One specific threat is the potential to generate toxic or harmful responses. Attackers can craft adversarial prompts that induce harmful responses from LLMs. In this work, we pioneer a theoretical foundation in LLMs security by identifying bias vulnerabilities within the safety fine-tuning and design a black-box jailbreak method named DRA (Disguise and Reconstruction Attack), which conceals harmful instructions through disguise and prompts the model to reconstruct the original harmful instruction within its completion. We evaluate DRA across various open-source and closed-source models, showcasing state-of-the-art jailbreak success rates and attack efficiency. Notably, DRA boasts a 91.1% attack success rate on OpenAI GPT-4 chatbot.

cs.CR↗

arXiv2024

AuthNet: Neural Network with Integrated Authentication Logic

Model stealing, i.e., unauthorized access and exfiltration of deep learning models, has become one of the major threats. Proprietary models may be protected by access controls and encryption. However, in reality, these measures can be compromised due to system breaches, query-based model extraction or a disgruntled insider. Security hardening of neural networks is also suffering from limits, for example, model watermarking is passive, cannot prevent the occurrence of piracy and not robust against transformations. To this end, we propose a native authentication mechanism, called AuthNet, which integrates authentication logic as part of the model without any additional structures. Our key insight is to reuse redundant neurons with low activation and embed authentication bits in an intermediate layer, called a gate layer. Then, AuthNet fine-tunes the layers after the gate layer to embed authentication logic so that only inputs with special secret key can trigger the correct logic of AuthNet. It exhibits two intuitive advantages. It provides the last line of defense, i.e., even being exfiltrated, the model is not usable as the adversary cannot generate valid inputs without the key. Moreover, the authentication logic is difficult to inspect and identify given millions or billions of neurons in the model. We theoretically demonstrate the high sensitivity of AuthNet to the secret key and its high confusion for unauthorized samples. AuthNet is compatible with any convolutional neural network, where our extensive evaluations show that AuthNet successfully achieves the goal in rejecting unauthenticated users (whose average accuracy drops to 22.03%) with a trivial accuracy decrease (1.18% on average) for legitimate users, and is robust against model transformation and adaptive attacks.

cs.CR↗

arXiv2024

SSL-WM: A Black-Box Watermarking Approach for Encoders Pre-trained by Self-supervised Learning

Recent years have witnessed tremendous success in Self-Supervised Learning (SSL), which has been widely utilized to facilitate various downstream tasks in Computer Vision (CV) and Natural Language Processing (NLP) domains. However, attackers may steal such SSL models and commercialize them for profit, making it crucial to verify the ownership of the SSL models. Most existing ownership protection solutions (e.g., backdoor-based watermarks) are designed for supervised learning models and cannot be used directly since they require that the models' downstream tasks and target labels be known and available during watermark embedding, which is not always possible in the domain of SSL. To address such a problem, especially when downstream tasks are diverse and unknown during watermark embedding, we propose a novel black-box watermarking solution, named SSL-WM, for verifying the ownership of SSL models. SSL-WM maps watermarked inputs of the protected encoders into an invariant representation space, which causes any downstream classifier to produce expected behavior, thus allowing the detection of embedded watermarks. We evaluate SSL-WM on numerous tasks, such as CV and NLP, using different SSL models both contrastive-based and generative-based. Experimental results demonstrate that SSL-WM can effectively verify the ownership of stolen SSL models in various downstream tasks. Furthermore, SSL-WM is robust against model fine-tuning, pruning, and input preprocessing attacks. Lastly, SSL-WM can also evade detection from evaluated watermark detection approaches, demonstrating its promising application in protecting the ownership of SSL models.

cs.CR↗

arXiv2023

DataElixir: Purifying Poisoned Dataset to Mitigate Backdoor Attacks via Diffusion Models

Dataset sanitization is a widely adopted proactive defense against poisoning-based backdoor attacks, aimed at filtering out and removing poisoned samples from training datasets. However, existing methods have shown limited efficacy in countering the ever-evolving trigger functions, and often leading to considerable degradation of benign accuracy. In this paper, we propose DataElixir, a novel sanitization approach tailored to purify poisoned datasets. We leverage diffusion models to eliminate trigger features and restore benign features, thereby turning the poisoned samples into benign ones. Specifically, with multiple iterations of the forward and reverse process, we extract intermediary images and their predicted labels for each sample in the original dataset. Then, we identify anomalous samples in terms of the presence of label transition of the intermediary images, detect the target label by quantifying distribution discrepancy, select their purified images considering pixel and feature distance, and determine their ground-truth labels by training a benign model. Experiments conducted on 9 popular attacks demonstrates that DataElixir effectively mitigates various complex attacks while exerting minimal impact on benign accuracy, surpassing the performance of baseline defense methods.

cs.CR↗

arXiv2023

Good-looking but Lacking Faithfulness: Understanding Local Explanation Methods through Trend-based Testing

While enjoying the great achievements brought by deep learning (DL), people are also worried about the decision made by DL models, since the high degree of non-linearity of DL models makes the decision extremely difficult to understand. Consequently, attacks such as adversarial attacks are easy to carry out, but difficult to detect and explain, which has led to a boom in the research on local explanation methods for explaining model decisions. In this paper, we evaluate the faithfulness of explanation methods and find that traditional tests on faithfulness encounter the random dominance problem, \ie, the random selection performs the best, especially for complex data. To further solve this problem, we propose three trend-based faithfulness tests and empirically demonstrate that the new trend tests can better assess faithfulness than traditional tests on image, natural language and security tasks. We implement the assessment system and evaluate ten popular explanation methods. Benefiting from the trend tests, we successfully assess the explanation methods on complex data for the first time, bringing unprecedented discoveries and inspiring future research. Downstream tasks also greatly benefit from the tests. For example, model debugging equipped with faithful explanation methods performs much better for detecting and correcting accuracy and security problems.

cs.LG↗