检索结果-内蒙古大学图书馆

Optimized Dynamic Watermarking for Audio DNNs with Adaptive Embedding and Boundary Sampling

学校读者我要写书评

暂无评论

Optimized Dynamic Watermarking for Audio DNNs with Adaptive ...

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Hao Fei Hewang Nie Siqi Sun Songfeng Lu Ting Luo Ling Qian Dunbo Cai Zhiguo Huang Runqing Zhang Hubei Key Laboratory of Distributed System Security Hubei Engineering Research Center on Big Data Security School of Cyber Science and Engineering Huazhong University of Science and Technology Wuhan China Shenzhen Huazhong University of Science and Technology Research Institute Shenzhen China China Mobile (Suzhou) Software Technology Company Limited Suzhou China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

The intensified concerns arising from the widespread adoption of deep learning have led to increased scrutiny of intellectual property protection in DNN models. Existing audio watermarking techniques, predominantly based on traditional signal processing methods, struggle to balance robustness, imperceptibility, and defense resistance in the face of evolving adversarial attacks. These limitations underscore the urgent need for more effective watermarking solutions in the audio domain. In this paper, we propose a dynamic audio watermarking framework that introduces an optimization-based approach to attach robust and adaptable triggers at arbitrary positions within audio signals, and innovatively integrates boundary sample selection driven by forgetting events and an adaptive watermark trigger embedding technique based on the SNR. Comprehensive experimental results reveal that our scheme preserves high model performance while maintaining remarkable stealthiness and robustness, offering a secure and reliable solution for safeguarding intellectual property in the audio domain and advancing the field of DNN watermarking.

关键词： Resistance Deep learning Adaptation models Watermarking Intellectual property Signal processing Robustness Speech processing Protection Signal to noise ratio

PicT: A SlimWeakly Supervised Vision Transformer for Pavement Distress Classification

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Tang, Wenhao Huang, Sheng Zhang, Xiaoxian Huangfu, Luwen School of Big Data & Software Engineering Chongqing University Chongqing China School of Big Data & Software Engineering Ministry of Education Key Laboratory of Dependable Service Computing in Cyber Physical Society Chongqing University Chongqing China Fowler College of Business Center for Human Dynamics in the Mobile Age San Diego State University San DiegoCA United States

Automatic pavement distress classification facilitates improving the efficiency of pavement maintenance and reducing the cost of labor and resources. A recently influential branch of this task divides the pavement image into patches and infers the patch labels for addressing these issues from the perspective of multi-instance learning. However, these methods neglect the correlation between patches and suffer from a low efficiency in the model optimization and inference. As a representative approach of vision Transformer, Swin Transformer is able to address both of these issues. It first provides a succinct and efficient framework for encoding the divided patches as visual tokens, then employs self-attention to model their relations. Built upon Swin Transformer, we present a novel vision Transformer named Pavement Image Classification Transformer (PicT) for pavement distress classification. In order to better exploit the discriminative information of pavement images at the patch level, the Patch Labeling Teacher is proposed to leverage a teacher model to dynamically generate pseudo labels of patches from image labels during each iteration, and guides the model to learn the discriminative features of patches via patch label inference in a weakly supervised manner. The broad classification head of Swin Transformer may dilute the discriminative features of distressed patches in the feature aggregation step due to the small distressed area ratio of the pavement image. To overcome this drawback, we present a Patch Refiner to cluster patches into different groups and only select the highest distress-risk group to yield a slim head for the final image classification. We evaluate our method on a large-scale bituminous pavement distress dataset named CQU-BPDD. Extensive results demonstrate the superiority of our method over baselines and also show that PicT outperforms the second-best performed model by a large margin of +2.4% in P@R on detection task, +3.9% in F 1 on recogni

关键词： Image classification

Securely Fine-tuning Pre-trained Encoders Against Adversarial Examples

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhou, Ziqi Li, Minghui Liu, Wei Hu, Shengshan Zhang, Yechao Wan, Wei Xue, Lulu Zhang, Leo Yu Yao, Dezhong Jin, Hai National Engineering Research Center for Big Data Technology and System China Services Computing Technology and System Lab. China Cluster and Grid Computing Lab. China Hubei Engineering Research Center on Big Data Security China Hubei Key Laboratory of Distributed System Security China School of Computer Science and Technology Huazhong University of Science and Technology China School of Software Engineering Huazhong University of Science and Technology China School of Cyber Science and Engineering Huazhong University of Science and Technology China School of Information and Communication Technology Griffith University Australia

With the evolution of self-supervised learning, the pre-training paradigm has emerged as a predominant solution within the deep learning landscape. Model providers furnish pre-trained encoders designed to function as versatile feature extractors, enabling downstream users to harness the benefits of expansive models with minimal effort through fine-tuning. Nevertheless, recent works have exposed a vulnerability in pre-trained encoders, highlighting their susceptibility to downstream-agnostic adversarial examples (DAEs) meticulously crafted by attackers. The lingering question pertains to the feasibility of fortifying the robustness of downstream models against DAEs, particularly in scenarios where the pre-trained encoders are publicly accessible to the attackers. In this paper, we initially delve into existing defensive mechanisms against adversarial examples within the pre-training paradigm. Our findings reveal that the failure of current defenses stems from the domain shift between pre-training data and downstream tasks, as well as the sensitivity of encoder parameters. In response to these challenges, we propose Genetic Evolution-Nurtured Adversarial Fine-tuning (Gen-AF), a two-stage adversarial fine-tuning approach aimed at enhancing the robustness of downstream models. Gen-AF employs a genetic-directed dual-track adversarial fine-tuning strategy in its first stage to effectively inherit the pre-trained encoder. This involves optimizing the pre-trained encoder and classifier separately while incorporating genetic regularization to preserve the model’s topology. In the second stage, Gen-AF assesses the robust sensitivity of each layer and creates a dictionary, based on which the top-k robust redundant layers are selected with the remaining layers held fixed. Upon this foundation, we conduct evolutionary adaptability fine-tuning to further enhance the model’s generalizability. Our extensive experiments, conducted across ten self-supervised training methods and six

关键词： Topology

Fed-AugMix: Balancing Privacy and Utility via data Augmentation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Haoyang Chen, Wei Zhang, Xiaojin National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Laboratory Cluster and Grid Computing Laboratory School of Computer Science and Technology Huazhong University of Science and Technology Wuhan430074 China School of Software Engineering Huazhong University of Science and Technology Wuhan430074 China

Gradient leakage attacks pose a significant threat to the privacy guarantees of federated learning. While distortion-based protection mechanisms are commonly employed to mitigate this issue, they often lead to notable performance degradation. Existing methods struggle to preserve model performance while ensuring privacy. To address this challenge, we propose a novel data augmentation-based framework designed to achieve a favorable privacy-utility trade-off, with the potential to enhance model performance in certain cases. Our framework incorporates the AugMix algorithm at the client level, enabling data augmentation with controllable severity. By integrating the Jensen-Shannon divergence into the loss function, we embed the distortion introduced by AugMix into the model gradients, effectively safeguarding privacy against deep leakage attacks. Moreover, the JS divergence promotes model consistency across different augmentations of the same image, enhancing both robustness and performance. Extensive experiments on benchmark datasets demonstrate the effectiveness and stability of our method in protecting privacy. Furthermore, our approach maintains, and in some cases improves, model performance, showcasing its ability to achieve a robust privacy-utility trade-off. Copyright © 2024, The Authors. All rights reserved.

关键词： Differential privacy

Edge computing service deployment and task offloading based on multi-task high-dimensional multi-objective optimization

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Guo, Yanheng Zhang, Yan Wu, Linjie Li, Mengxia Cai, Xingjuan Chen, Jinjun Shanxi Key Laboratory of Big Data Analysis and Parallel Computing Taiyuan University of Science and Technology Shanxi Taiyuan030024 China State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Software Engineering Swinburne University of Technology Melbourne Australia

The Mobile Edge Computing (MEC) system located close to the client allows mobile smart devices to offload their computations onto edge servers, enabling them to benefit from low-latency computing services. Both cloud service providers and users seek more comprehensive solutions, necessitating judicious decisions in service deployment and task offloading while balancing multiple objectives. This study investigates service deployment and task offloading challenges in a multi-user environment, framing them as a multi-task high-dimensional multi-objective optimization (MT-HD-MOO) problem within an edge environment. To ensure stable service provisioning, beyond considering latency, energy consumption, and cost as deployment objectives, network reliability is also incorporated. Furthermore, to promote equitable usage of edge servers, load balancing is introduced as a fourth task offloading objective, in addition to latency, energy consumption, and cost. Additionally, this paper designs a MT-HD-MOO algorithm based on a multi-selection strategy to address this model and its solution. By employing diverse selection strategies, an environment selection strategy pool is established to enhance population diversity within the high-dimensional objective space. Ultimately, the algorithm’s effectiveness is verified through simulation experiments. © 2023, CC BY-NC-ND.

关键词： Multiobjective optimization

EG-KGR: A Knowledge Graph Reasoning Model Based on Enhanced Graph Sample and Aggregate Inductive Learning Algorithm

学校读者我要写书评

暂无评论

EG-KGR: A Knowledge Graph Reasoning Model Based on Enhanced ...

International Conference on Tools for Artificial Intelligence (ICTAI)

作者： Yuejia Wu Jian-Tao Zhou Inner Mongolia Key Laboratory of Discipline Inspection and Supervision Big Data Inner Mongolia Engineering Laboratory for Big Data Analysis Technology Ministry of Education Inner Mongolia Engineering Laboratory for Cloud Computing and Service Software Inner Mongolia Key Laboratory of Social Computing and Data Processing College of Computer Science Inner Mongolia University National & Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian Engineering Research Center of Ecological Big Data Hohhot China

Knowledge Graph is an important research field that involves the storage and management of knowledge, but the incompleteness and sparsity of Knowledge Graphs hinder their application in many fields. Knowledge Graph Reasoning aims to alleviate this problem by completing missing paths or identifying wrong paths between entities. Graph Convolution Network (GCN) based methods are one of the state-of-the-art approaches to this work. However, it is difficult to directly generalize to unknown nodes and utilizes valid information from the local neighborhood which results in poor flexibility and extensibility and will loss of important information. This paper presents EG-KGR, a plug-and-play knowledge reasoning model based on enhanced graph sampling and aggregate inductive learning algorithm to relieve the above problems and enhance existing GCN-based methods. Specifically, EG-KGR supports incremental characteristics, uses inductive learning to replace transductive learning, and designs random sampling and local information sampling optimization methods to improve the model's generalization ability, prediction accuracy, and running speed. Extensive experimental results show that our EG-KGR can achieve optimal results.

关键词： Training Knowledge engineering Convolution Aggregates Optimization methods Knowledge graphs Learning (artificial intelligence)

Fence Theorem: Towards Dual-Objective Semantic-Structure Isolation in Preprocessing Phase for 3D Anomaly Detection

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Liang, Hanzhe Zhou, Jie Chen, Xuanxin Dai, Tao Wang, Jinbao Gao, Can College of Computer Science and Software Engineering Shenzhen University China Shenzhen Audencia Financial Technology Institute Shenzhen University China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University China Faculty of Education Shenzhen University China Guangdong Provincial Key Laboratory of Intelligent Information Processing China

3D anomaly detection (AD) is prominent but difficult due to lacking a unified theoretical foundation for preprocessing design. We establish the Fence Theorem, formalizing preprocessing as a dual-objective semantic isolator: (1) mitigating cross-semantic interference to the greatest extent feasible and (2) confining anomaly judgements to aligned semantic spaces wherever viable, thereby establishing intra-semantic comparability. Any preprocessing approach achieves this goal through a two-stage process of Semantic-Division and Spatial-Constraints stage. Through systematic deconstruction, we theoretically and experimentally subsume existing preprocessing methods under this theorem via tripartite evidence: qualitative analyses, quantitative studies, and mathematical proofs. Guided by the Fence Theorem, we implement Patch3D, consisting of Patch-Cutting and Patch-Matching modules, to segment semantic spaces and consolidate similar ones while independently modeling normal features within each space. Experiments on Anomaly-ShapeNet and Real3D-AD with different settings demonstrate that progressively finer-grained semantic alignment in preprocessing directly enhances point-level AD accuracy, providing inverse validation of the theorem’s causal logic. Copyright © 2025, The Authors. All rights reserved.

关键词： Fences

Category-Prompt Refined Feature Learning for Long-Tailed Multi-Label Image Classification

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yan, Jiexuan Huang, Sheng Mu, NanKun Huangfu, Luwen Liu, Bo School of Big Data & Software Engineering Chongqing University Chongqing China College of Computer Science Chongqing University Chongqing China Fowler College of Business San Diego State University San DiegoCA United States School of Computer Science and Information Engineering Hefei University of Technology Hefei China

Real-world data consistently exhibits a long-tailed distribution, often spanning multiple categories. This complexity underscores the challenge of content comprehension, particularly in scenarios requiring Long-Tailed Multi-Label image Classification (LTMLC). In such contexts, imbalanced data distribution and multi-object recognition pose significant hurdles. To address this issue, we propose a novel and effective approach for LTMLC, termed Category-Prompt Refined Feature Learning (CPRFL), utilizing semantic correlations between different categories and decoupling category-specific visual representations for each category. Specifically, CPRFL initializes category-prompts from the pretrained CLIP’s embeddings and decouples category-specific visual representations through interaction with visual features, thereby facilitating the establishment of semantic correlations between the head and tail classes. To mitigate the visual-semantic domain bias, we design a progressive Dual-Path Back-Propagation mechanism to refine the prompts by progressively incorporating context-related visual information into prompts. Simultaneously, the refinement process facilitates the progressive purification of the category-specific visual representations under the guidance of the refined prompts. Furthermore, taking into account the negative-positive sample imbalance, we adopt the Asymmetric Loss as our optimization objective to suppress negative samples across all classes and potentially enhance the head-to-tail recognition performance. We validate the effectiveness of our method on two LTMLC benchmarks and extensive experiments demonstrate the superiority of our work over baselines. The code is available at https://***/jiexuanyan/CPRFL. © 2024, CC BY-NC-SA.

关键词： Semantics

Dual Latent State Learning: Exploiting Regional Network Similarities for QoS Prediction

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wang, Ziliang Zhang, Xiaohong Zhang, Kechi Li, Ze Shi Yan, Meng Ministry of Education School of Computer Science Peking University Beijing China Ministry of Education China School of Big Data and Software Engineering Chongqing University Chongqing401331 China The Department of Computer Science the University of Victoria Canada

Individual objects (users or services) within a region usually have similar network states to each other because they usually come from the same region. Despite the similarity within regional networks, many existing techniques overlook this potential, thereby limiting the accuracy of Quality of Service (QoS) predictions. In this paper, we introduce the regional-based dual latent state learning network(R2SL), a novel deep learning framework designed to overcome the pitfalls of traditional individual object-based prediction techniques for QoS prediction. R2SL first captures the regional network states by deriving two distinct regional network latent states: the physical-regional latent state and the virtual-regional latent state. These states are built using aggregated data from a physical area (such as the same city) or a virtual network area (such as the same service provider) rather than individual object data. Then R2SL provides a specific Mixture of Experts(MOE) network based on sparse activation to realize the classification learning task of latent features of different regions. Thirdly, R2SL adopts an enhanced smooth Huber loss function that adjusts Huber loss’s linear loss component to alleviate the impact caused by label imbalance. Finally, the perceptual network was used to interpret the integrated features, thereby realizing the QoS prediction. Experimental results show that compared with the typical methods and the latest methods, the proposed method can obtain higher prediction accuracy, thus improving the research level of QoS prediction. © 2023, CC0.

关键词： Contrastive Learning