检索结果-内蒙古大学图书馆

FSSA: Efficient 3-Round Secure Aggregation for Privacy-Preserving Federated Learning

学校读者我要写书评

暂无评论

TechRxiv

TechRxiv 2022年

作者： Luo, Fucai Al-Kuwari, Saif Yan, Xingfu Wang, Haiyan The Department of New Networks Peng Cheng Laboratory Shenzhen China The College of Science and Engineering Hamad Bin Khalifa University Doha Qatar The Guangdong Provincial Key School of Computer Science and Engineering South China University of Technology Guangzhou China

Federated learning (FL) allows a large number of clients to collaboratively train machine learning (ML) models by sending only their local gradients to a central server for aggregation in each training iteration, without sending their raw training data. Unfortunately, recent attacks on FL demonstrate that local gradients may leak information about local training data. In response to such attacks, Bonawitz et al. (CCS 2017) proposed a secure aggregation protocol that allows a server to compute the sum of clients’ local gradients in a secure manner. However, their secure aggregation protocol requires at least 4 rounds of communication between each client and the server in each training iteration. The number of communication rounds is closely related not only to the total communication cost but also the ML model accuracy, as the number of communication rounds affects client dropouts. In this paper, we propose FSSA, a 3-round secure aggregation protocol, that is efficient in terms of computation and communication, and resilient to client dropouts. We prove the security of FSSA in honest-but-curious setting and show that the security can be maintained even if an arbitrarily chosen subset of clients drop out at any time. We evaluate the performance of FSSA and show that its computation and communication overhead remains low even on large datasets. Furthermore, we conduct an experimental comparison between FSSA and Bonawitz et al.’s protocol. The comparison results show that, in addition to reducing the number of communication rounds, FSSA achieves a significant improvement in computational efficiency. © 2022, CC BY.

关键词： Machine learning

Robust Reinforcement Learning for Risk-Sensitive Linear Quadratic Gaussian Control

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Cui, Leilei Başar, Tamer Jiang, Zhong-Ping The Control and Networks Lab Department of Electrical and Computer Engineering Tandon School of Engineering New York University BrooklynNY11201 United States The Coordinated Science Laboratory University of Illinois Urbana-Champaign UrbanaIL61801 United States

This paper proposes a novel robust reinforcement learning framework for discrete-time linear systems with model mismatch that may arise from the sim-to-real gap. A key strategy is to invoke advanced techniques from control theory. Using the formulation of the classical risk-sensitive linear quadratic Gaussian control, a dual-loop policy optimization algorithm is proposed to generate a robust optimal controller. The dual-loop policy optimization algorithm is shown to be globally and uniformly convergent, and robust against disturbances during the learning process. This robustness property is called small-disturbance input-to-state stability and guarantees that the proposed policy optimization algorithm converges to a small neighborhood of the optimal controller as long as the disturbance at each learning step is relatively small. In addition, when the system dynamics is unknown, a novel model-free off-policy policy optimization algorithm is proposed. Finally, numerical examples are provided to illustrate the proposed algorithm. Copyright © 2022, The Authors. All rights reserved.

关键词： Reinforcement learning

SBPF: Sensitiveness Based Pruning Framework For Convolutional Neural Network On Image Classification

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Lu, Yiheng Gong, Maoguo Zhao, Wei Feng, Kai-Yuan Li, Hao State Key Laboratory of Integrated Services Networks School of Computer Science and Technology Xidian University Xi’an710071 China Key Laboratory of Intelligent Perception and Image Understanding School of Electronic Engineering Ministry of Education Xidian University Xi’an710071 China

Pruning techniques are used comprehensively to compress convolutional neural networks (CNNs) on image classification. However, the majority of pruning methods require a well pre-trained model to provide useful supporting parameters, such as C1-norm, BatchNorm value and gradient information, which may lead to inconsistency of filter evaluation if the parameters of the pre-trained model are not well optimized. Therefore, we propose a sensitiveness based method to evaluate the importance of each layer from the perspective of inference accuracy by adding extra damage for the original model. Because the performance of the accuracy is determined by the distribution of parameters across all layers rather than individual parameter, the sensitiveness based method will be robust to update of parameters. Namely, we can obtain similar importance evaluation of each convolutional layer between the imperfect-trained and fully trained models. For VGG-16 on CIFAR-10, even when the original model is only trained with 50 epochs, we can get same evaluation of layer importance as the results when the model is trained fully. Then we will remove filters proportional from each layer by the quantified sensitiveness. Our sensitiveness based pruning framework is verified efficiently on VGG-16, a customized Conv-4 and ResNet-18 with CIFAR-10, MNIST and CIFAR-100, respectively. Copyright © 2022, The Authors. All rights reserved.

关键词： Convolutional neural networks

An embarrassingly simple approach to semi-supervised few-shot learning 22

学校读者我要写书评

暂无评论

An embarrassingly simple approach to semi-supervised few-sho...

Proceedings of the 36th International Conference on Neural Information Processing Systems

作者： Xiu-Shen Wei He-Yang Xu Faen Zhang Yuxin Peng Wei Zhou School of Computer Science and Engineering Nanjing University of Science and Technology and State Key Laboratory of Integrated Services Networks Xidian University School of Computer Science and Engineering Nanjing University of Science and Technology Qingdao AInnovation Technology Group Co. Ltd Wangxuan Institute of Computer Technology Peking University CICC Alpha (Beijing) Private Equity

ISBN: (纸本)9781713871088

Semi-supervised few-shot learning consists in training a classifier to adapt to new tasks with limited labeled data and a fixed quantity of unlabeled data. Many sophisticated methods have been developed to address the challenges this problem comprises. In this paper, we propose a simple but quite effective approach to predict accurate negative pseudo-labels of unlabeled data from an indirect learning perspective, and then augment the extremely label-constrained support set in few-shot classification tasks. Our approach can be implemented in just few lines of code by only using off-the-shelf operations, yet it is able to outperform state-of-the-art methods on four benchmark datasets.

关键词：

A Framework Based on Generational and Environmental Response Strategies for Dynamic Multi-objective Optimization

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Qingya Liu, Xiangzhi Wang, Fuqiang Wang, Shuai Zhang, Peng Wu, Xiaoming Guangdong Provincial Key Laboratory of Brain-Inspired Intelligent Computation Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen518055 China Shandong Provincial Key Laboratory of Computer Networks Shandong China Heze Branch Qilu University of Technology Shandong Academy of Sciences Biological Engineering Technology Innovation Center of Shandong Province Shandong China

Due to the dynamics and uncertainty of the dynamic multi-objective optimization problems (DMOPs), it is difficult for algorithms to find a satisfactory solution set before the next environmental change, especially for some complex environments. One reason may be that the information in the environmental static stage can not be used well in the traditional framework. In this paper, a novel framework based on generational and environmental response strategies (FGERS) is proposed, in which response strategies are run both in the environmental change stage and the environmental static stage to obtain population evolution information of those both stages. Unlike in the traditional framework, response strategies are only run in the environmental change stage. For simplicity, the feed-forward center point strategy was chosen to be the response strategy in the novel dynamic framework (FGERS-CPS). FGERS-CPS is not only to predict change trend of the optimum solution set in the environmental change stage, but to predict the evolution trend of the population after several generations in the environmental static stage. Together with the feed-forward center point strategy, a simple memory strategy and adaptive diversity maintenance strategy were used to form the complete FGERS-CPS. On 13 DMOPs with various characteristics, FGERS-CPS was compared with four classical response strategies in the traditional framework. Experimental results show that FGERS-CPS is effective for DMOPs. Copyright © 2022, The Authors. All rights reserved.

关键词： Genetic algorithms

Data Identification Optimal Control for Quantized Linear Systems under DoS Attacks

学校读者我要写书评

暂无评论

IEEE Transactions on Aerospace and Electronic Systems 2025年

作者： Wang, Xianming Li, Liwei Shen, Mouquan Zhao, Xudong Wang, Qing-Guo Zhu, Zheng Hong Nanjing Tech University School of Mechanical and Power Engineering Nanjing211816 China Nanjing Tech University College of Electrical Engineering and Control Science Nanjing211816 China Dalian University of Technology Faculty of Electronic Information and Electrical Engineering Dalian116024 China Beijing Normal University Institute of Ai and Future Networks Zhuhai519087 China BNU-HKBU United International College Guangdong Provincial Key Laboratory Irads Department of Computer Science Zhuhai519087 China York University Department of Mechanical Engineering TorontoONM3J 1P3 Canada

This paper is devoted to data identification optimal control for quantized linear systems under denial of service (DoS) attacks. A least-squares algorithm is provided to identify the unknown system matrices using a data-based representation. An output estimation algorithm utilizing identified matrices is supplied to offset DoS attacks. Consequently, an optimal controller is built on the quantized and attacked data by adaptive dynamic programming. Sufficient conditions are presented to guarantee that the resultant closed-loop system is ultimately uniformly bounded. An example is produced to verify the validity of the proposed approach. © 1965-2011 IEEE.

关键词： Optimal control systems

Secure the 5G and Beyond networks with Zero Trust and Access Control Systems for Cloud Native Architectures

学校读者我要写书评

暂无评论

Secure the 5G and Beyond Networks with Zero Trust and Access...

ACS/IEEE International Conference on computer Systems and Applications

作者： Hisham A. Kholidy Keven Disen Andrew Karam Elhadj Benkhelifa Mohammad A. Rahman Atta-Ur Rahman Ibrahim Almazyad Ahmed F. Sayed Rakia Jaziri Dept. of Networks and Computer Security College of Engineering SUNY Polytechnic Institute Utica NY USA Dept. of Networks and Comp. Security SUNY Polytechnic Institute Utica NY USA The Air Force Research Laboratory (AFRL RIGB) Rome NY USA Smart Systems AI and Cybersecurity Research Centre Staffordshire Univ. UK Dept of Electrical and Computer Engineering Florida International University USA Dept. of Computer Science Imam Abdulrahman Bin Faisal University Dammam Saudi Arabia Dept. of Electrical and Computer Engineering Univ. of Arizona Tucson AZ USA Transmission Dept. Telecom Egypt Fayoum Egypt Paragraphe Research Lab. Univ. of Paris VIII France

5G networks are highly distributed, built on an open service-based architecture that requires multi-vendor hardware and software development environments, all of which create a high attack surface in the 5G networks than other proprietary fixed-function networks. Besides that, cloud-native architectures also present new security challenges. Cloud-native separates monolithic virtual machines into microservice pods, resulting in higher volumes of signaling and communication flowing through and between microservices. In addition, secure connections in monolithic applications have been replaced by untrusted communication between microservice pods, requiring additional cybersecurity capabilities. Access control systems were created to provide reliability and limit access to an organization’s assets. However, due to technology's constant evolution and dynamicity, these conventional security systems lack the security to protect an organization’s information because they were created to address access control for known users. For 5G based cloud native technology, these access controls need to be taken further by implementing a Zero Trust model to secure one’s essential assets for all users within the system. Zero Trust is implemented in an access control system under the concept "Never Trust, Always Verify". In this paper, we implement zero trust as a factor within access control systems by combining the principles of access control systems and zero-trust security by factoring in the user’s historical behavior and recommendations into the mix.

关键词：

Enhancing Security and Efficiency in Vehicle-to-Sensor Authentication: A Multi-Factor Approach with Cloud Assistance 23

学校读者我要写书评

暂无评论

Enhancing Security and Efficiency in Vehicle-to-Sensor Authe...

23rd IEEE International Conference on Trust, Security and Privacy in Computing and Communications, TrustCom 2024

作者： Zhang, Xinrui Zhao, Pincan Jaskolka, Jason Carleton University Department of Systems and Computer Engineering OttawaK1S 5B6 Canada Xidian University State Key Laboratory of Integrated Services Networks Xi'an710071 China Carleton University School of Information Technology OttawaK1S 5B6 Canada

Connected and Autonomous Vehicles (CAVs) can improve their perception by integrating data from roadside sensors. However, ensuring secure authentication between CAVs and sensors is challenging due to the limited capabilities of sensors and the growing number of vehicles. This paper introduces a secure authentication protocol that enables direct communication between CAVs and roadside sensors, addressing a critical gap in existing research focused on vehicle-to-cloud authentication. The proposed multi-factor authentication scheme combines password, biometric, and device-specific factors with Elliptic Curve Cryptography (ECC) and efficient key agreement protocols. A comprehensive adversary model tailored for vehicular networks is presented, along with an in-depth security analysis demonstrating the scheme's resilience against various threats. The cloud-assisted authentication framework offloads computationally intensive tasks to the cloud server, reducing the burden on resource-constrained Roadside Units (RSUs) and ensuring scalability. Extensive performance evaluations showcase the scheme's computational efficiency, low communication overhead, and storage costs compared to state-of-the-art solutions, highlighting its practical feasibility and potential for real-world deployment in intelligent transportation systems. © 2024 IEEE.

关键词： Authentication Protocol

Unlocking Mode Programming with Multi-Plane Light Conversion Using computer-Generated Hologram Optimisation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Rothe, Stefan Barbosa, Fabio Czarske, Jürgen W. Ferreira, Filipe M. Department of Applied Physics Yale University New HavenCT06520 United States Optical Networks Dept Electronic & Electrical Eng University College London United Kingdom TU Dresden Faculty of Electrical and Computer Engineering Laboratory of Measurement and Sensor System Technique Dresden01062 Germany TU Dresden Faculty of Physics Dresden01062 Germany

Programmable optical devices provide performance enhancement and flexibility to spatial multiplexing systems enabling transmission of tributaries in high-order eigenmodes of spatially-diverse transmission media, like multimode fibre (MMF). Wavefront shaping with spatial light modulators (SLMs) facilitates scalability of the transmission media by allowing for channel diagonalization and quasi-single-input single-output operation. Programmable mode multiplexing configurations like multi-plane light conversion (MPLC) utilise the SLM and offer the potential to simultaneously launch an arbitrary subset of spatial tributaries in any N-mode MMF. Such programmable optical processor would enable the throughput of space-division multiplexing (SDM) systems to be progressively increased by addressing a growing number of tributaries over one MMF and in this way meet a growing traffic demand – similarly to the wavelength-division multiplexing evolution path. Conventionally, MPLC phasemasks are calculated using the wavefront matching algorithm (WMA). However, this method does not exploit the full potential of programmable mode multiplexers. We show, that computer-generated hologram algorithms like direct search enable significant improvement compared to the traditional WMA-approach. Such gains are enabled by tailored cost functions with dynamic constraints concerning insertion loss as well as mode extinction ratio. We show that average mode extinction ratio can be greatly improved by as much as ≈ 15 dB at the expense of insertion loss deterioration of © 2024, CC BY.

关键词： Light modulators