MXMingfu Xue

Mingfu Xue

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Mingfu Xue has a total of 58 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • GNNLicense: An Active Intellectual Property Protection Technique for Graph Neural Networks based on Autoencoder

    by Mingfu Xue, Longlong Hu, Linji Zhang, Yang Shi, Liquan Chen

    Graph Neural Networks (GNNs) are widely utilized in areas such as social network analysis, recommendation systems, and fraud detection. Protecting the intellectual property (IP) of GNN models is crucial to avoid potential economic losses. However, existing researches on GNN IP protection rely on watermarking techniques for passive copyright verification. In this paper, we propose GNNLicense, the first active GNN IP protection and user’s identity management method against GNN piracy. The proposed method employs encoder-decoder models to encode and reconstruct graph features. Since the graph encoded by the encoder can only be decoded by the corresponding decoder, this design achieves active IP protection. In addition, the proposed method leverages fingerprint to manage and distinguish between different authorized users’ identities. The encoder is held by the user, while the decoder is retained by the service provider. The user encodes the input graph data through the encoder, and then submits it to the server for normal predictions. This mechanism makes unauthorized users can only obtain very poor model performance while ensures the normal use of authorized users. Experimental results on five graph datasets and five popular GNN models across node classification and graph classification tasks demonstrate the method’s effectiveness. Moreover, the method maintains its robustness when subjected to fine-tuning and adaptive attacks.


  • IsBad: Imperceptible Sample-Specific Backdoor to DNN With Denoising Autoencoder

    by Xiangqi Wang, Mingfu Xue, Kewei Chen, Jing Xu, Wenmao Liu, Leo Yu Zhang, Yushu Zhang

    The backdoor attack poses a new security threat to deep neural networks (DNNs). The existing backdoor often relies on visible universal triggers to make the backdoored model malfunction, which is not only usually visually suspicious to humans but also catchable by mainstream countermeasures. We propose an imperceptible sample-specific backdoor that the trigger varies from sample to sample and is invisible. Our trigger generation is automated through a denoising autoencoder that is fed with delicate but pervasive features (e.g., edge patterns per image). We extensively experiment with our backdoor attack on ImageNet and MS-Celeb-1 M, which demonstrates a stable and nearly 100% (i.e., 99.8%) attack success rate with negligible impact on the clean data accuracy of the infected model. The denoising autoencoder-based trigger generator is reusable or transferable across tasks (e.g., from ImageNet to MS-Celeb-1 M), whilst the trigger has high exclusiveness (i.e., a trigger generated for one sample is not applicable to another sample). Besides, our proposal backdoored model has achieved high evasiveness against mainstream backdoor defenses such as Neural Cleanse, STRIP, SentiNet, Fine-Pruning, BDMAE, and BTI-DBF.


  • An Active Authorization Control Method for Deep Reinforcement Learning Model Based on GANs and Adaptive Trigger

    by Mingfu Xue, Kewei Chen, Leo Yu Zhang, Yushu Zhang, Weiqiang Liu

    In recent years, deep reinforcement learning (DRL) has found widespread applications across diverse scenarios. Since the DRL training process requires substantial time and financial costs, well-trained DRL policies should be considered as intellectual property (IP) which deserves proper protection. However, to date, there are only a few studies on IP protection on DRL and the existing methods are limited to passive copyright verification. In this paper, we propose the first active authorization control method for DRL which can proactively protect deep reinforcement learning policy. The DRL policy trained with this method can be used by authorized users normally, but cannot be used by unauthorized users (i.e., the protected policy’s performance for unauthorized users is paralyzed). Specifically, we train a trigger injection network and a discriminator network based on generative adversarial networks (GANs). During the DRL policy training phase, we use trigger injection network to insert sample-specific triggers to all observations and use triggered observations to train the protected policy. Our approach is applicable across various deep reinforcement learning algorithms. We conduct effectiveness experiments on different DRL policies trained using different DRL algorithms, and the experimental results revealed that the performance of authorized users is on par with the performance of clean DRL policy trained normally (baseline), whereas the performance of unauthorized users significantly deviates from that of the baseline. Specifically, the authorized performance of protected Breakout-DQN, Breakout-A2C, MsPacman-DQN and MsPacman-A2C policies are 416.4 (baseline 397.8), 403.0 (baseline 415.0), 2552.0 (baseline 2472.0), and 1964.0 (baseline 1828.0). Comparatively, the unauthorized performance of protected Breakout-DQN, Breakout-A2C, MsPacman-DQN and MsPacman-A2C policies are only 4.4 (baseline 397.8), 2.0 (baseline 415.0), 74.0 (baseline 2472.0), and 514.0 (baseline 1828.0). Furthermore, the experiments demonstrate that the proposed method exhibits robustness against pruning, fine-tuning, and adaptive attacks.


  • 3D Invisible Cloak: A Robust Person Stealth Attack Against Object Detector in Complex 3D Physical Scenarios

    by Mingfu Xue, Can He, Yushu Zhang, Zhe Liu, Weiqiang Liu

    In this paper, we propose a novel physical stealth attack against the person detectors in real world. For the first time, we consider the impacts of those complex and challenging 3D physical constraints ( e.g. , radian, wrinkle, occlusion, angle, etc .) on person stealth attacks, and propose 3D transformations to generate robust 3D invisible cloak. We launch the person stealth attacks in 3D physical space instead of 2D plane by printing the adversarial patches on real clothes. Anyone wearing the cloak can evade the detection of person detectors and achieve stealth under challenging and complex 3D physical scenarios. Experimental results in various indoor and outdoor physical scenarios show that, the proposed person stealth attack method is robust and effective even under those complex and challenging physical conditions, such as the cloak is wrinkled, obscured, curved, and from different/large angles. The attack success rate of the generated adversarial patch in digital domain (Inria data set) is 86.56% against YOLO v2 and 80.32% against YOLO v5, while the static and dynamic stealth attack success rates of the generated 3D invisible cloak in physical world are 100%, 77% against YOLO v2 and 100%, 83.95% against YOLO v5, respectively, which are significantly better than state-of-the-art works.


  • Open Access

    SSAT: Active Authorization Control and User’s Fingerprint Tracking Framework for DNN IP Protection

    by Mingfu Xue, Yinghao Wu, Leo Yu Zhang, Dujuan Gu, Yushu Zhang, Weiqiang Liu

    As training a high-performance deep neural network (DNN) model requires a large amount of data, powerful computing resources and expert knowledge, protecting well-trained DNN models from intellectual property (IP) infringement has raised serious concerns in recent years. Most existing methods using DNN watermarks to verify the ownership of the models after IP infringement occurs, which is reactive in the sense that they cannot prevent unauthorized users from using the model in the first place. Different from these methods, in this article, we propose an active authorization control and user’s fingerprint tracking method for the IP protection of DNN models by utilizing sample-specific backdoor attack. The proposed method inversely and multiplely exploits sample-specific trigger as the key to implement authorization control for DNN model, in which the generated triggers are imperceptible and sample-specific for clean images. Specifically, a U-Net model is used to generate backdoor instances. Then, the target model is trained on the clean images and backdoor instances, which are inversely labeled as wrong classes and correct classes, respectively. Only authorized users can use the target model normally by pre-processing the clean images through the U-Net model. Moreover, the images processed by the U-Net model will contain unique fingerprint that can be extracted to verify and track the corresponding user’s identity. This article is the first work that utilizes the sample-specific backdoor attack to implement active authorization control and user’s fingerprint management for DNN model under black-box scenarios. Extensive experimental results on ImageNet dataset and YouTube Aligned Face dataset demonstrate that the proposed method is effective in protecting the DNN model from unauthorized usage. Specifically, the protected model has a low inference accuracy (1.00%) for unauthorized users, while maintaining a normal inference accuracy (97.67%) for authorized users. Besides, the proposed method can achieve 100% fingerprint tracking success rates on both the ImageNet and YouTube Aligned Face datasets. Moreover, it is demonstrated that the proposed method is robust against fine-tuning attack, pruning attack, pruning attack with retraining, reverse-engineering attack, adaptive attack, and JPEG compression attack. The code is available at https://github.com/nuaaaisec/SSAT .


  • A Reinforcement Learning-Based ELF Adversarial Malicious Sample Generation Method

    by Mingfu Xue, Jinlong Fu, Zhiyuan Li, Shifeng Ni, Heyi Wu, Leo Yu Zhang, Yushu Zhang, Weiqiang Liu

    In recent years, domestic Linux operating systems have developed rapidly, but the threat of ELF viruses has become increasingly prominent. Currently, domestic antivirus software for information technology application innovation (ITAI) operating systems shows insufficient capability in detecting ELF viruses. At the same time, research on generating malicious samples in ELF format is scarce. In order to fill this gap at home and abroad and meet the growing application needs of domestic antivirus software companies, this paper proposes an automatic ELF adversarial malicious samples generation technique based on reinforcement learning. Based on reinforcement learning framework, after being processed by cycles of feature extraction, malicious detection, agent decision-making, and evade-detection operation, the sample can evade the detection of antivirus engines. Specifically, nine feature extractor subclasses are used to extract features in multiple aspects. The PPO algorithm is used as the agent algorithm. The action table in the evade-detection module contains 11 evade-detection operations for ELF malicious samples. This method is experimentally verified on the ITAI operating system, and the ELF malicious sample set on the Linux x86 platform is used as the original sample set. The detection rate of this sample set by ClamAV before processing is 98%, and the detection rate drops to 25% after processing. The detection rate of this sample set by 360 Security before processing is 4%, and the detection rate drops to 1% after processing. Furthermore, after processing, the average number of engines on VirusTotal that could detect the maliciousness of the sample decreases from 39 to 15. Many malicious samples were detected by 41~43 engines on VirusTotal before processing, while after the evade-detection processing, only 8~9 engines on VirusTotal can detect the malware. In terms of executability and malicious function consistency, the processed samples can still run normally and the malicious functions remain consistent with those before processing. Overall, the proposed method in this paper can effectively generate adversarial ELF malware samples. Using this method to generate malicious samples to test and train the anti-virus software can promote and improve anti-virus software’s detection and defense capability against malware.


  • Open Access

    An Imperceptible and Owner-unique Watermarking Method for Graph Neural Networks

    by Linji Zhang, Mingfu Xue, Leo Yu Zhang, Yushu Zhang, Weiqiang Liu

    Graph Neural Networks (GNNs) have found widespread application across various domains, encompassing but not limited to social network analysis, recommendation systems, and fraud detection. Meanwhile, training a sophisticated GNN model is an extremely resource-intensive process. Therefore, protecting the intellectual property of GNN model becomes essential. However, limited research has been conducted on the protection of intellectual property for GNNs. Additionally, current few watermarking methods employed in the context of GNNs overlook the potential vulnerabilities posed by evasion attack and fraudulent declaration attack. To fill this gap, in this paper, we propose a novel GNN watermarking method utilizing a bi-level optimization framework to embed an imperceptible and owner-unique watermark into GNNs. The proposed method achieves indistinguishability and uniqueness of the injected watermark, establishing a secure mechanism for intellectual property protection for GNNs. We evaluate our method on two benchmark datasets and three GNN models. The results indicate that our method effectively verifies model ownership with minimal impact on their primary task performance. Furthermore, the method exhibits remarkable resilience against model fine-tuning and pruning attacks, as well as security against evasion attacks and fraudulent ownership claims.


  • Open Access

    Tracking the Leaker: An Encodable Watermarking Method for Dataset Intellectual Property Protection

    by Yifan Shang, Mingfu Xue, Leo Yu Zhang, Yushu Zhang, Weiqiang Liu

    Presently, numerous enterprises provide machine learning cloud services. However, the service provider may exploit user-uploaded data for unauthorized model retraining or illicit collection of user data for commercial model development. This study introduces a traceable dataset watermarking technique designed to ascertain the trustworthiness of third-party providers offering machine learning cloud services. In the event of a data breach, the source can be traced back to the suspicious third-party responsible for data leakage. Specifically, we propose a method that employs the clean-label backdoor attack framework to infer whether a third-party model is trained using user data. A watermark, associated with the encoding and designed as a trigger, is injected into the dataset through a trained autoencoder. Experimental evaluation on three datasets proves the effectiveness of the proposed method, yielding over 93% accuracy on average under normal conditions. A series of pruning and fine-tuning attacks were carried out on the method, with the results indicating that these attacks have a minimal impact and confirming the method’s robustness.


  • An Explainable Intellectual Property Protection Method for Deep Neural Networks Based on Intrinsic Features

    by Mingfu Xue, Xin Wang, Yinghao Wu, Shifeng Ni, Leo Yu Zhang, Yushu Zhang, Weiqiang Liu

    Intellectual property (IP) protection for Deep Neural Networks (DNNs) has raised serious concerns in recent years. Most existing works embed watermarks in the DNN model for IP protection, which need to modify the model and do not consider/mention interpretability. In this paper, for the first time, we propose an interpretable intellectual property protection method for DNN based on explainable artificial intelligence. Compared with existing works, the proposed method does not modify the DNN model, and the decision of the ownership verification is interpretable. We extract the intrinsic features of the DNN model by using Deep Taylor Decomposition. Since the intrinsic feature is composed of unique interpretation of the model’s decision, the intrinsic feature can be regarded as fingerprint of the model. If the fingerprint of a suspected model is the same as the original model, the suspected model is considered as a pirated model. Experimental results demonstrate that the fingerprints can be successfully used to verify the ownership of the model and the test accuracy of the model is not affected. Furthermore, the proposed method is robust to fine-tuning attack, pruning attack, watermark overwriting attack, and adaptive attack.


  • Imperceptible and multi-channel backdoor attack

    by Mingfu Xue, Shifeng Ni, Yinghao Wu, Yushu Zhang, Weiqiang Liu