检索结果-内蒙古大学图书馆

DarkFed: A data-Free Backdoor Attack in Federated Learning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Minghui Wan, Wei Ning, Yuxuan Hu, Shengshan Xue, Lulu Zhang, Leo Yu Wang, Yichen School of Software Engineering Huazhong University of Science and Technology China National Engineering Research Center for Big Data Technology and System China Services Computing Technology and System Lab China Hubei Engineering Research Center on Big Data Security China Hubei Key Laboratory of Distributed System Security China School of Cyber Science and Engineering Huazhong University of Science and Technology China School of Computer Science and Technology Huazhong University of Science and Technology China School of Information and Communication Technology Griffith University Australia

Federated learning (FL) has been demonstrated to be susceptible to backdoor attacks. However, existing academic studies on FL backdoor attacks rely on a high proportion of real clients with main task-related data, which is impractical. In the context of real-world industrial scenarios, even the simplest defense suffices to defend against the state-of-the-art attack, 3DFed. A practical FL backdoor attack remains in a nascent stage of development. To bridge this gap, we present DarkFed. Initially, we emulate a series of fake clients, thereby achieving the attacker proportion typical of academic research scenarios. Given that these emulated fake clients lack genuine training data, we further propose a data-free approach to backdoor FL. Specifically, we delve into the feasibility of injecting a backdoor using a shadow dataset. Our exploration reveals that impressive attack performance can be achieved, even when there is a substantial gap between the shadow dataset and the main task dataset. This holds true even when employing synthetic data devoid of any semantic information as the shadow dataset. Subsequently, we strategically construct a series of covert backdoor updates in an optimized manner, mimicking the properties of benign updates, to evade detection by defenses. A substantial body of empirical evidence validates the tangible effectiveness of DarkFed. Copyright © 2024, The Authors. All rights reserved.

关键词： Semantics

eSROP Attack: Leveraging Signal Handler to Implement Turing-Complete Attack Under CFI Defense 1

学校读者我要写书评

暂无评论

18th EAI International Conference on Security and Privacy in Communication Networks, SecureComm 2022

作者： Zhang, Tianning Cai, Miao Zhang, Diming Huang, Hao Department of Computer Science and Technology Nanjing University Nanjing China State Key Laboratory for Novel Software Technology Nanjing University Nanjing China Key Laboratory of Water Big Data Technology of Ministry of Water Resources Hohai University Nanjing China School of Computer and Information Hohai University Nanjing China College of Computer Engineering Jiangsu University of Science and Technology Nanjing China

ISBN: (数字)9783031255380

ISBN: (纸本)9783031255373

Signal Return Oriented Programming (SROP) is a dangerous code reuse attack method. Recently, defense techniques have been proposed to defeat SROP attacks. In this paper, we leverage the signal nesting mechanism provided by current operating systems and propose a new variant of SROP attack called enhanced SROP (eSROP) attack. eSROP provides the ability of invoking arbitrary system calls, simulating Turing-complete computation, and even bypassing the fine-grained label-based CFI defense, without modifying the return address and instruction register in the signal frame. Because the signal returns to the interrupted instruction, the shadow stack defense can hardly detect our attack. Signal has strong flexibility which can interrupt the normal control flow. We leverage such flexibility to design a new code reuse attack. To evaluate eSROP, we perform two exploits on two real-world programs, namely Proftpd and Wu-ftpd. In our attacks, adversaries can invoke arbitrary system calls and obtain a root shell. Both attacks succeed within 10 min under strict system defense such as data execution prevention, address space layout randomization, and coarse-grained control flow integrity. © 2023, ICST Institute for Computer Sciences, Social Informatics and Telecommunications engineering.

关键词： Network security

Enhance Learning Efficiency of Oblique Decision Tree via Feature Concatenation

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Lyu, Shen-Huan He, Yi-Xiao Wang, Yanyan Qu, Zhihao Tang, Bin Ye, Baoliu Key Laboratory of Water Big Data Technology of Ministry of Water Resources College of Computer Science and Software Engineering Hohai University Nanjing China State Key Laboratory for Novel Software Technology Nanjing University Nanjing China School of Artificial Intelligence and Information Technology Nanjing University of Chinese Medicine Nanjing China

Oblique Decision Tree (ODT) separates the feature space by linear projections, as opposed to the conventional Decision Tree (DT) that forces axis-parallel splits. ODT has been proven to have a stronger representation ability than DT, as it provides a way to create shallower tree structures while still approximating complex decision boundaries. However, its learning efficiency is still insufficient, since the linear projections cannot be transmitted to the child nodes, resulting in a waste of model parameters. In this work, we propose an enhanced ODT method with Feature Concatenation (FC-ODT), which enables in-model feature transformation to transmit the projections along the decision paths. Theoretically, we prove that our method enjoys a faster consistency rate w.r.t. the tree depth, indicating that our method possesses a significant advantage in generalization performance, especially for shallow trees. Experiments show that FC-ODT can outperform the other state-of-the-art decision trees with a limited tree depth. Copyright © 2025, The Authors. All rights reserved.

关键词： Efficiency

BADROBOT: JAILBREAKING EMBODIED LLMS IN THE PHYSICAL WORLD

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhang, Hangtao Zhu, Chenyu Wang, Xianlong Zhou, Ziqi Yin, Changgan Li, Minghui Xue, Lulu Wang, Yichen Hu, Shengshan Liu, Aishan Guo, Peijin Zhang, Leo Yu 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 Beihang University China School of Cyber Science and Engineering Huazhong University of Science and Technology China School of Software Engineering Huazhong University of Science and Technology China School of Information and Communication Technology Griffith University Australia

Embodied AI represents systems where AI is integrated into physical entities. Large Language Model (LLM), which exhibits powerful language understanding abilities, has been extensively employed in embodied AI by facilitating sophisticated task planning. However, a critical safety issue remains overlooked: could these embodied LLMs perpetrate harmful behaviors? In response, we introduce BADROBOT, a novel attack paradigm aiming to make embodied LLMs violate safety and ethical constraints through typical voice-based user-system interactions. Specifically, three vulnerabilities are exploited to achieve this type of attack: (i) manipulation of LLMs within robotic systems, (ii) misalignment between linguistic outputs and physical actions, and (iii) unintentional hazardous behaviors caused by world knowledge’s flaws. Furthermore, we construct a benchmark of various malicious physical action queries to evaluate BADROBOT’s attack performance. Based on this benchmark, extensive experiments against existing prominent embodied LLM frameworks (e.g., Voxposer, Code as Policies, and ProgPrompt) demonstrate the effectiveness of our BADROBOT. Copyright © 2024, The Authors. All rights reserved.

关键词： Benchmarking

Breaking Barriers in Physical-World Adversarial Examples: Improving Robustness and Transferability via Robust Feature

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Yichen Chou, Yuxuan Zhou, Ziqi Zhang, Hangtao Wan, Wei Hu, Shengshan Li, Minghui 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 Cyber Science and Engineering Huazhong University of Science and Technology China School of Computer Science and Technology Huazhong University of Science and Technology China School of School of Software Engineering Huazhong University of Science and Technology China

As deep neural networks (DNNs) are widely applied in the physical world, many researches are focusing on physical-world adversarial examples (PAEs), which introduce perturbations to inputs and cause the model’s incorrect outputs. However, existing PAEs face two challenges: unsatisfactory attack performance (i.e., poor transferability and insufficient robustness to environment conditions), and difficulty in balancing attack effectiveness with stealthiness, where better attack effectiveness often makes PAEs more perceptible. In this paper, we explore a novel perturbation-based method to overcome the challenges. For the first challenge, we introduce a strategy Deceptive RF injection based on robust features (RFs) that are predictive, robust to perturbations, and consistent across different models. Specifically, it improves the transferability and robustness of PAEs by covering RFs of other classes onto the predictive features in clean images. For the second challenge, we introduce another strategy Adversarial Semantic Pattern Minimization, which removes most perturbations and retains only essential adversarial patterns in AEs. Based on the two strategies, we design our method Robust Feature Coverage Attack (RFCoA), comprising Robust Feature Disentanglement and Adversarial Feature Fusion. In the first stage, we extract target class RFs in feature space. In the second stage, we use attention-based feature fusion to overlay these RFs onto predictive features of clean images and remove unnecessary perturbations. Experiments show our method’s superior transferability, robustness, and stealthiness compared to existing state-of-the-art methods. Additionally, our method’s effectiveness can extend to Large Vision-Language Models (LVLMs), indicating its potential applicability to more complex tasks. Copyright © 2024, The Authors. All rights reserved.

关键词： Deep neural networks

A Study Of Voiceprint Recognition Technology Based on Deep Learning

学校读者我要写书评

暂无评论

A Study Of Voiceprint Recognition Technology Based on Deep L...

International Wireless Communications and Mobile Computing Conference, IWCMC

作者： Jingyi Li Qin Xu Michel Kadoch School of Big Data and Software Chongqing College of Mobile Telecommunications Chongqing China Department of Electrical Engineering ETS University of Quebec Canada

Voiceprint recognition technology is a high-quality identification technology, which is currently widely used in all walks of life and brings great convenience to our lives. In order to better promote the actual application range and effects of voiceprint recognition technology, the research and realization of its technology has extremely far-reaching significance. This paper mainly analyzes the basic process of voiceprint recognition, technical indicators, fac-tors that affect the effect of voiceprint recognition, and the current research status of voiceprint recognition technology. In addition, it analyzes the ad-vantages of the current mature deep learning technology, as well as the main problems existing in the current voiceprint recognition research work, and gives directions for the next step in the effective combination of deep learning technology for voiceprint recognition research work.

关键词： Deep learning Wireless communication Speech recognition Spectrogram Mobile computing

An improved conflict evidence management approach using base belief function for uncertain prior information modeling

学校读者我要写书评

暂无评论

An improved conflict evidence management approach using base...

第42届中国控制会议

作者： Yongchuan Tang Yonghao Zhou Ying Zhou Shuang Ni Yubo Huang Deyun Zhou School of Microelectronics Northwestern Polytechnical University School of Computer Science Northwestern Polytechnical University School of Electronics and Information Northwestern Polytechnical University School of Big Data and Software Engineering Chongqing University School of Engineering University of Warwick

Due to the complexity in practical environment,there are a variety of interference resulting in inaccurate information which will directly affect the final result of data *** the era of big data,as the number of fusing information sources increases,even if they are consistent,conflicts may ***,classical Dempster combination rule in DempsterShafer evidence theory(D-S theory) cannot solve the problem of conflict data ***,an improved method is proposed for conflict data fusion by assigning a base belief to each piece of *** this paper,the base belief function is used to construct the initial belief degree ***,the belief entropy is calculated to get the information volume of each *** combination rule is used to get the final result after evidence modification,which can help solve conflict data fusion *** examples and experiments are used to verity the rationality and effectiveness of the proposed method.

关键词：

Geographic-Semantic-Temporal Hypergraph Convolutional Network for Traffic Flow Prediction

学校读者我要写书评

暂无评论

Geographic-Semantic-Temporal Hypergraph Convolutional Networ...

International Conference on Pattern Recognition

作者： Kesu Wang Jing Chen Shijie Liao Jiaxin Hou Qingyu Xiong School of Big Data & Software Engineering Chongqing University Chongqing China

ISBN: (纸本)9781728188089;9781728188096

Traffic flow forecasting has become an increasingly important part of intelligent traffic control and management. This task is challenging due to (1) complex geographic and non-geographic spatial correlations; (2) temporal correlations between time slices; (3) dynamics of semantic high-order correlations along the temporal dimension. To address those difficulties, commonly-used methods apply graph convolutional networks for spatial correlations and recurrent neural networks for temporal dependencies. In this work, we distinguish the two aspects of spatial correlations and propose the two types of spatial graphs, named as geographic graph and semantic hypergraph. Furthermore, we extend the traditional convolution into geographic-temporal graph convolution and semantic-temporal hypergraph convolution to jointly capture geographic-temporal correlations and semantic-temporal correlations. Then we propose a geographic-semantic-temporal hypergraph convolutional network (GST-HCN) that combines our graph convolutions and GRU units hierarchically in a unified end-to-end network. The experiment results on the Caltrans Performance Measurement System (PeMS) dataset show that our proposed model significantly outperforms other popular spatio-temporal deep learning models and suggests the effectiveness to explore geographic-semantic-temporal dependencies on deep learning models for traffic flow prediction.

关键词： Deep learning Measurement Correlation Recurrent neural networks Convolution Semantics Predictive models

Dependency-Aware Hybrid Task Offloading in Mobile Edge Computing Networks 27

学校读者我要写书评

暂无评论

Dependency-Aware Hybrid Task Offloading in Mobile Edge Compu...

27th IEEE International Conference on Parallel and Distributed Systems, ICPADS 2021

作者： Ming, Zhao Li, Xiuhua Sun, Chuan Fan, Qilin Wang, Xiaofei Leung, Victor C. M. School of Big Data and Software Engineering Chongqing University Chongqing China TKLAN College of Intelligence and Computing Tianjin University Tianjin China College of Computer Science and Software Engineering Shenzhen University Shenzhen China The University of British Columbia Department of Electrical and Computer Engineering Vancouver Canada

ISBN: (纸本)9781665408783

With the rapid increase of data in mobile edge computing (MEC) networks, mobile devices (MDs) have been generating many computation-latency-sensitive tasks. As the MDs are limited by resources in terms of storage, computation, and bandwidth, part of tasks have to be offloaded to the edge of mobile networks or the remote cloud for more efficient processing. Hence, task offloading plays a vital role in this scene. Existing works about task offloading mainly aim at one-shot task offloading and rarely consider the dependencies of tasks. In this paper, we focus on minimizing the maximum delay of processing a series of tasks with dependencies in MEC networks, which supports device-to-device communications. Specifically, we consider task offloading under a hybrid scenario with a small base station (SBS) deployed with an edge server (ES) and several MDs which generate several tasks with dependencies. Then we model the tasks to a weighted directed acyclic graph (DAG) and formulate the optimization problem as minimizing the critical path of the weighted DAG. To tackle this NP-hard problem, we propose a heuristic scheme to iteratively optimize the delay of paths of the weighted DAG under the constraints of the ES. To evaluate the proposed scheme, we perform numerical experiments with different numbers of tasks. Simulation results demonstrate that the proposed scheme outperforms other schemes in terms of reducing the system delay and saving the energy consumption of the MDs. © 2021 IEEE.

关键词： Optimization