Yuhang Zhao

Yuhang Zhao

Auto-generated profile, based on publicly available information

Yuhang Zhao

School of Cyberspace Science and Technology, Beijing Institute of Technology, Beijing, China

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 August 2021 in International Journal of Intelligent Systems
Verified
Journal Article
Closed access
PDF file
YMYuxi Ma
Yuxi Ma
MSMeng Shen
Meng Shen
YZYuhang Zhao
Yuhang Zhao
ZLZhao Li
Zhao Li
XTXiaoyao Tong
Xiaoyao Tong
Zhi WangZhi Wang
Zhi Wang

14 November 2024 in IEEE Transactions on Cognitive Communications and Networking
Verified
Journal Article
Closed access
YZYuhang Zhao
Yuhang Zhao
Yajie WangYajie Wang
Yajie Wang
Chuan ZhangChuan Zhang
Chuan Zhang
CLChunhai Li
Chunhai Li
ZXZehui Xiong
Zehui Xiong
DNDusit Niyato
Dusit Niyato
In the radio frequency field, deep neural networks have been widely used for automatic modulation recognition tasks due to their superior accuracy. However, it has been shown that these models are susceptible to adversarial examples, which are the kinds of carefully crafted perturbations that can lead to model misclassification and raise security issues in applications. To solve this problem, we propose an Ultra-Fusion Adversarial Training method, which combines adversarial training and ensemble learning to enable the model robustness to withstand different attack strengths. We explore the number and distribution of ensembled attacks and introduce a Fermi-function-like distribution to optimally balance the performance of different attack strengths. Additionally, we investigate the effect of the signal-to-noise ratio (SNR) interval on the model accuracy and robustness, suggesting the effective SNR interval for training. Considering the demand for practical application scenarios of modulation recognition, we propose a comprehensive robustness metric based on weighted integral to evaluate the robustness of the trained models. Numerical experiments demonstrate that our method improves the model’s robustness by 31.89% against white-box attacks, and achieves up to an 80.54% improvement in black-box scenarios. These results show that our method has the ability to resiliently resist potential attacks of various strengths and can be applied to spectrum application scenarios with high-security requirements.

6 March 2023 in IEEE Transactions on Artificial Intelligence
Verified
Journal Article
Closed access
HZHuipeng Zhou
Huipeng Zhou
Yajie WangYajie Wang
Yajie Wang
YTYu-An Tan
Yu-An Tan
SWShangbo Wu
Shangbo Wu
YZYuhang Zhao
Yuhang Zhao
YLYuanzhang Li
Yuanzhang Li
In recent years, deep neural networks (DNNs) have made significant progress on face recognition. However, DNNs have been found to be vulnerable to adversarial examples, leading to fatal consequences in real-world applications. This paper focuses on improving the transferability of adversarial examples against face recognition models. We propose Gradient Eroding (GE) to make the gradient of the residual blocks more diverse, by eroding the back-propagation dynamically. We also propose a novel black-box adversarial attack named Corrasion Attack based on GE. Extensive experiments demonstrate that our approach can effectively improve the transferability of adversarial attacks against face recognition models. Our approach over-performs 29.35% in fooling rate than state-of-the-art black-box attacks. Leveraging adversarial training with adversarial examples generated by us, the robustness of models can be improved by up to 43.2%. Besides, Corrasion Attack successfully breaks two online face recognition systems, achieving a highest fooling rate of 89.8%.

12 December 2024 in IEEE Transactions on Cognitive Communications and Networking
Verified
Journal Article
Closed access
YZYuhang Zhao
Yuhang Zhao
Zuobin YingZuobin Ying
Zuobin Ying
Yajie WangYajie Wang
Yajie Wang
Chuan ZhangChuan Zhang
Chuan Zhang
JZJun Zheng
Jun Zheng
Liehuang ZhuLiehuang Zhu
Liehuang Zhu
Automatic modulation recognition (AMR) plays a crucial role in wireless communications. Deep learning-based AMR methods have garnered significant attention due to their high accuracy. Among these, transformer-based models hold great promise but often underperform in noisy environments. To address this problem, we propose a specific transformer model tailored for signal data, named radio transformer (RaT), which aims to extract frequency-domain features to achieve higher classification accuracy. We design a frequency encoder that contains an adaptive filter to assign weights to each frequency component. Additionally, we introduce global average pooling (GAP) in the classification stage to reduce the impact of noise. Furthermore, we analyze the model’s representational capability with varying slice lengths and encoder layers, identifying the optimal configurations for both parameters. Experimental results on the RML 2018.01A dataset demonstrate that our proposed RaT model achieves superior performance, particularly on low signal-to-noise (SNR) conditions. The accuracy is improved by up to 15.79% when the SNR is 4 dB, underscoring its effectiveness and potential for real-time communication applications.

4 November 2019
Verified
Book Chapter
Closed access
PDF file
YZYuhang Zhao
Yuhang Zhao
KWKunqing Wang
Kunqing Wang
YXYuan Xue
Yuan Xue
QZQuanxin Zhang
Quanxin Zhang
XZXiaosong Zhang
Xiaosong Zhang