Search

Your search keyword '"Gao, Yuanzhe"' showing total 162 results
Start Over Author "Gao, Yuanzhe"
162 results on '"Gao, Yuanzhe"'

1. DMPA: Model Poisoning Attacks on Decentralized Federated Learning for Model Differences

Author

Feng, Chao, Li, Yunlong, Gao, Yuanzhe, Celdrán, Alberto Huertas, von der Assen, Jan, Bovet, Gérôme, and Stiller, Burkhard

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Federated learning (FL) has garnered significant attention as a prominent privacy-preserving Machine Learning (ML) paradigm. Decentralized FL (DFL) eschews traditional FL's centralized server architecture, enhancing the system's robustness and scalability. However, these advantages of DFL also create new vulnerabilities for malicious participants to execute adversarial attacks, especially model poisoning attacks. In model poisoning attacks, malicious participants aim to diminish the performance of benign models by creating and disseminating the compromised model. Existing research on model poisoning attacks has predominantly concentrated on undermining global models within the Centralized FL (CFL) paradigm, while there needs to be more research in DFL. To fill the research gap, this paper proposes an innovative model poisoning attack called DMPA. This attack calculates the differential characteristics of multiple malicious client models and obtains the most effective poisoning strategy, thereby orchestrating a collusive attack by multiple participants. The effectiveness of this attack is validated across multiple datasets, with results indicating that the DMPA approach consistently surpasses existing state-of-the-art FL model poisoning attack strategies., Comment: 8 pages, 3 figures

Published
2025

2. From Models to Network Topologies: A Topology Inference Attack in Decentralized Federated Learning

Author

Feng, Chao, Gao, Yuanzhe, Celdran, Alberto Huertas, Bovet, Gerome, and Stiller, Burkhard

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Federated Learning (FL) is widely recognized as a privacy-preserving machine learning paradigm due to its model-sharing mechanism that avoids direct data exchange. However, model training inevitably leaves exploitable traces that can be used to infer sensitive information. In Decentralized FL (DFL), the overlay topology significantly influences its models' convergence, robustness, and security. This study explores the feasibility of inferring the overlay topology of DFL systems based solely on model behavior, introducing a novel Topology Inference Attack. A taxonomy of topology inference attacks is proposed, categorizing them by the attacker's capabilities and knowledge. Practical attack strategies are developed for different scenarios, and quantitative experiments are conducted to identify key factors influencing the attack effectiveness. Experimental results demonstrate that analyzing only the public models of individual nodes can accurately infer the DFL topology, underscoring the risk of sensitive information leakage in DFL systems. This finding offers valuable insights for improving privacy preservation in decentralized learning environments.

Published
2025

7. Novel Topology Inference Attacks on DFL

Author

Gao, Yuanzhe and Gao, Yuanzhe

Abstract

Decentralized Federated Learning (DFL), a new paradigm in the field of machine learning, reaches the goal of collaboratively training advanced models while preserving local data privacy by allowing multiple participants to forward model parameters to each other. Due to its decentralized nature, the network topology can be diverse. Different topologies have a significant impact on the performance and security of DFL. Therefore, this work proposes a novel topology inference attack specifically for DFL networks. This attack can be divided into two scenarios based on the level of information the attacker has: (1) In the first scenario, the attacker has all the node information and some edge information, and predicts the connection status of unknown edges by training a supervised machine learning model; (2) In the second scenario, the attacker only has all the node information and determines whether different nodes in the network are connected by using a clustering model from unsupervised learning framework. In addition, this work constructs four effective DFL network node metrics to facilitate these topology inference attacks. The designed attack methods of this work have been evaluated in various DFL environments and achieves good performance in most of them, thus providing new insights for future securiy research in the field of DFL network.

Published
2024

34. Preparation of LDO@TiO2 core–shell nanosheets for enhanced photocatalytic degradation of organic pollution.

Author

Wang, Can, Zhang, Ruikang, Miao, Yucong, Xue, Qihui, Yu, Borong, Gao, Yuanzhe, Han, Zhangang, and Shao, Mingfei

Subjects
NANOSTRUCTURED materials, LAYERED double hydroxides, CORE materials, POLLUTION, DISCONTINUOUS precipitation, PHOTODEGRADATION
Abstract

TiO2-based nanosheet materials with a core–shell structure are expected to be one of the promising photocatalysts for the degradation of organic pollution. It is a challenge to synthesize TiO2 by the desired nucleation and growth process, so most reported TiO2 core–shell photocatalysts are prepared using TiO2 as a core material. Layered double hydroxides (LDHs) are considered ideal platforms to grow TiO2in situ and further serve as additional components to improve the separation of photogenerated charge carriers. In this work, we report the design and fabrication of anatase TiO2-coated ZnAl-layered double oxide (LDO@TiO2) nanosheets, which involve the in situ growth of TiO2 on ZnAl-LDH followed by subsequent calcination treatment. The resulting LDO@TiO2 photocatalyst yields typical core–shell nanosheet morphology with a mesoporous structure, exhibiting excellent photodegradation and mineralization efficiency for organic pollution. [ABSTRACT FROM AUTHOR]

Published
2021
Full Text
View/download PDF

41. Impact of oxygen vacancies on TiO2 charge carrier transfer for photoelectrochemical water splitting.

Author

Huang, Xiaoqian, Gao, Xiaoru, Xue, Qihui, Wang, Can, Zhang, Ruikang, Gao, Yuanzhe, and Han, Zhangang

Subjects
CHARGE carriers, CHARGE transfer, WATER transfer, METAL oxide semiconductors, PHOTOCATHODES, OXYGEN
Abstract

Oxygen vacancies are recognized as the most prevalent defects in oxide materials. The effect of oxygen vacancies on the physicochemical properties of metal oxide semiconductors has attracted considerable attention in the photocatalysis field. But so far, the impact of oxygen vacancies on charge carrier transfer for photoelectrochemical water splitting has been unclear. In this work, TiO2 photoanodes with various oxygen vacancy concentrations were studied as metal oxide models to clarify the impact of oxygen vacancies on charge carrier transfer behaviors. The potential distribution and electrochemical impedance spectroscopy results indicate that the oxygen vacancies facilitate charge carrier diffusion in TiO2, but are disadvantageous for the charge carrier drift in the TiO2/electrolyte interface. The TiO2-400 photoanode with intermediate oxygen vacancy concentration exhibits the highest photocurrent density. It is expected that this work will provide reference to design and fabricate oxide semiconductors as photoanodes for higher charge carrier utilization in the field of solar-to-chemical energy conversion. [ABSTRACT FROM AUTHOR]

Published
2020
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results