Conditional Generative Adversarial networks (cGANs) are increasingly popular web-based synthesis services accessed through a query API, e.g., cGANs generate a cat image based on a "cat" query. However, cGAN-...
详细信息
ISBN:
(纸本)9798331530044;9798331530037
Conditional Generative Adversarial networks (cGANs) are increasingly popular web-based synthesis services accessed through a query API, e.g., cGANs generate a cat image based on a "cat" query. However, cGAN-based synthesizers can be stolen via adversaries' queries, i.e., model thieves. The prevailing adversarial assumption is that thieves act independently: they query the deployed cGAN (i.e., the victim), and train a stolen cGAN using the images obtained from the victim. A popular anti-theft defense consists in throttling down the number of queries from any given user. We consider a more realistic adversarial scenario: model thieves collude to query the victim, and then train the stolen cGAN. CLUES is a new collusive model stealing framework, enabling thieves to bypass throttle-based defenses and steal cGANs more efficiently than through individual efforts. Thieves collect queried images and train a stolen cGAN in a federated manner. We evaluate CLUES on three image datasets, e.g., MNIST, FashionMNIST and CelebA. We experimentally show the scalability of the proposed attack strategies against the number of thieves and the queried images, the impact of a classical noise-based defense, a passive watermarking defense and a JPEG-based countermeasure. Our evaluation shows that such a collusive stealing strategy gets close to 4 units of Frechet Inception Distance from a victim model. Our code is readily available to the research community: https://***/records/10224340.
In the study of signal estimation, the performance of distributed blind equalization over wireless sensor network (WSN) is significantly affected by channel conditions. However, there has been few studies investigatin...
详细信息
Cellular networks, renowned for their robustness and high availability, must consistently meet very stringent standards to ensure service provision and attract deployment in manufacturing industries that cannot tolera...
详细信息
ISBN:
(纸本)9798350362527;9798350362510
Cellular networks, renowned for their robustness and high availability, must consistently meet very stringent standards to ensure service provision and attract deployment in manufacturing industries that cannot tolerate any downtime. Despite the progressive introduction of numerous algorithms over the years across different layers of the 5G communication model to ensure packet delivery and decoding, jamming attacks have consistently disrupted connectivity. Detection and analysis of jamming attack characteristics, including their type and duration, provide comprehensive insights for security analysts to develop educated countermeasures. This paper demonstrates how computer vision can be employed on spectrograms to detect, classify, and log information on three types of jamming attacks in real-time within a 5G campus network. We evaluate the efficiency of our real-time jamming detector by analyzing detection latency, classification accuracy, and attack duration capture accuracy. Our results indicate that our Deep Learning classification model detects all tested jammers with remarkably high accuracy.
As semiconductor design approaches physical limits, computer processing speeds are stagnating. This poses significant challenges for traffic simulations, which are becoming more and more computationally demanding. To ...
详细信息
ISBN:
(纸本)9798350369205;9798350369199
As semiconductor design approaches physical limits, computer processing speeds are stagnating. This poses significant challenges for traffic simulations, which are becoming more and more computationally demanding. To maintain fast execution times while accommodating more complex simulations, it is essential to utilize the parallel computing capabilities of modern hardware. This paper discusses the need for an updated architectural design in the MATSim traffic simulation framework to take advantage of parallel computing infrastructures. We introduce a prototype that adapts the existing traffic simulation logic to a distributed parallel algorithm. Extensive benchmarks have been conducted to evaluate the prototype's performance and identify its limitations. The results demonstrate that the prototype performs up to 100 times faster than the current implementation. Based on these findings, we advocate for the integration of a distributed traffic simulation within the MATSim framework and outline necessary steps to enhance the prototype.
Developing efficient load-balancing techniques remains a persistent research challenge as modern WiFi networks evolve into increasingly complex environments, incorporating new enhancements in their standards. For inst...
详细信息
ISBN:
(纸本)9798331531317;9798331531300
Developing efficient load-balancing techniques remains a persistent research challenge as modern WiFi networks evolve into increasingly complex environments, incorporating new enhancements in their standards. For instance, DQL-MultiMDP is a load-balancing algorithm that learns an optimal STA-to-AP association policy to ensure user fairness and optimize network performance in dense and dynamic WiFi networks. The algorithm leverages a Multi-Markov Decision Process (Multi-MDP) strategy to accommodate the fluctuating number of devices caused by their switching on/off. However, scalability challenges arise due to the exponential expansion of the action space. In this paper, we propose a divide-and-conquer approach that extends the algorithm to operate in extremely large deployments: a dynamic partitioning mechanism divides the network into clusters and assigns a sub-controller to manage the STA-to-AP association in each cluster independently, and a coordination mechanism enables them to exchange their training updates. Experimental investigations validate the effectiveness of the approach and motivate future work.
This paper delves into the potential of leveraging federated learning (FL) techniques to revolutionize vehicle-to-everything (V2X) wireless communications. We examine the significant opportunities that FL offers in th...
详细信息
ISBN:
(纸本)9798350364279;9798350364262
This paper delves into the potential of leveraging federated learning (FL) techniques to revolutionize vehicle-to-everything (V2X) wireless communications. We examine the significant opportunities that FL offers in the V2X domain, including distributed model training, privacy preservation and enhanced communication efficiency. Concurrently, we spotlight the formidable challenges associated with adapting FL to the unique characteristics of highly dynamic and heterogeneous vehicular networks. These challenges encompass Non Independent Identically Distribution (Non-IID) data distribution, ensuring model reliability and addressing scoring disparities among participating vehicles. To navigate this intricate landscape, we propose novel research directions, including hierarchical FL, incentive mechanisms and the development of robust training algorithms, aimed at surmounting these obstacles. This paper draws upon valuable insights gleaned from an initial analysis, along with simulation results that underscore the efficacy of tailored FL designs in unlocking the transformative potential of collective learning within vehicular networks, while simultaneously mitigating systemic limitations.
The integration of distributed generators (DGs) poses great challenges to the topology, operation planning, control methods, and protection configuration of traditional distribution networks (DN). To solve the optimiz...
详细信息
This study investigates the impact of role fixation (RF) and role rotation (RR) scripts on learners' socially shared regulation (SSR) in computer-supported collaborative learning (CSCL) environments. The research ...
详细信息
ISBN:
(纸本)9798350361414
This study investigates the impact of role fixation (RF) and role rotation (RR) scripts on learners' socially shared regulation (SSR) in computer-supported collaborative learning (CSCL) environments. The research involved 25 undergraduate students from a university in central China, divided into RF and RR groups for an eight-week collaborative learning activity focused on the computernetworks course. Data were collected through self-report questionnaires and online discussions platform. Findings indicate that RF groups exhibited significantly higher levels of SSR and more balanced discussions compared to RR groups. RF groups fostered a more evenly distributed interaction network, enhancing cooperation and sharing. These results underscore the effectiveness of role fixation scripts in promoting SSR and improving collaborative learning outcomes.
distributed Validator Technology (DVT) mitigates single points of failure in Ethereum validators by distributing validator duties across a node cluster and making decisions within the cluster through consensus algorit...
详细信息
ISBN:
(纸本)9798350348439;9798350384611
distributed Validator Technology (DVT) mitigates single points of failure in Ethereum validators by distributing validator duties across a node cluster and making decisions within the cluster through consensus algorithms. However, these systems often fail to ensure economic fairness for participants, where misconduct by a few nodes can lead to financial losses for honest nodes within the cluster. To address this issue, we introduce RobustETH, an innovative DVT implementation that prioritizes operational efficiency and economic fairness for participants. Specifically, RobustETH incorporates an efficient consensus protocol, X-IBFT, to enhance DVT's operational efficiency. Additionally, we propose a BFT forensic protocol and a reputation-based recluster strategy to accurately penalize malicious nodes and mitigate their impact, thereby safeguarding the financial interests of honest participants. Our theoretical analysis proves RobustETH's safety, liveness, and accountability. Moreover, through comprehensive evaluations, we demonstrate that RobustETH achieves a 58% reduction in end-to-end latency compared to current state-of-the-art DVT implementations while ensuring economic fairness.
Digital Inline Holographic Microscopy (DIHM) is a lensless shadow imaging technique where a coherent light-source is utilized to illuminate samples and record interference patterns in a digitalizing device. Those patt...
详细信息
ISBN:
(纸本)9798350384734;9798350384727
Digital Inline Holographic Microscopy (DIHM) is a lensless shadow imaging technique where a coherent light-source is utilized to illuminate samples and record interference patterns in a digitalizing device. Those patterns, called holograms, carry volumetric information regarding the inspected sample, requiring complex numerical methods to make them distinguishable as occurs in a conventional bright-field microscopic image. However, for many applications, the real-time analysis of holograms is computationally expensive due to the nature of numerical diffraction methods to reconstruct the signals into visual information. To mitigate this problem, in this paper we investigate the use of deep learning approaches to classify those interference patterns, directly from the raw holograms, without the requirement of phase-recovering methods for diffraction. In our approach, we investigated the use of distinct Convolutional Neural networks (CNNs) architectures and its adaptability to correctly classify holograms in an experimental environment with a dataset generated from synthetic interference patterns produced by the Fresnel Diffraction Method. The computer-generated dataset was produced from 26 classes, resulting in a total of 520 samples after the data augmentation procedure. The obtained results demonstrated the feasibility of the proposed approach to properly classify samples with 96.8% of precision, directly from the holographic interference patterns, avoiding the need for computationally expensive diffraction methods.
暂无评论