Yuchi Yao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yuchi Yao has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Haojie Ji
Key Laboratory of Modern Measurement & Control Technology, Ministry of Education, Beijing Information Science & Technology University, Beijing 100192, China
2 shared publications - Chongshi Xin
Beijing Information Science and Technology University
1 shared publication - Jiangyue Wang
Key Laboratory of Modern Measurement & Control Technology, Ministry of Education, Beijing Information Science & Technology University, Beijing 100192, China
1 shared publication - Jingyan Wang
Key Laboratory of Modern Measurement & Control Technology, Ministry of Education, Beijing Information Science & Technology University, Beijing 100192, China
1 shared publication - Junjie Zhang
Beihang University
1 shared publication - Liangliang Tian
Key Laboratory of Modern Measurement & Control Technology, Ministry of Education, Beijing Information Science & Technology University, Beijing 100192, China
1 shared publication
Latest Publications
- Open Access
Face Privacy Protection Method for Autonomous Sensors Based on Hierarchical Format-Preserving Encryption
Advanced sensors in connected automated vehicles (CAVs) increasingly collect facial biometric information for environmental perception, posing serious privacy leakage risks. However, existing privacy protection methods for automotive data primarily focus on strict security mechanisms and fail to fully balance data usability. This paper presents a hierarchical format-preserving encryption (H-FPE) method for face privacy protection in autonomous sensors. The proposed method constructs a privacy-preserving framework for face detection based on YOLOv11 by employing a region-specific encryption strategy where the encryption strength is tailored to the importance of different facial regions. The encryption algorithm employs SM4-based Feistel structures with pseudo-random functions to ensure RGB value constraints while maintaining image format integrity. Experimental evaluation results in diverse scenarios demonstrate that the proposed privacy encryption method achieves superior privacy protection performance. In terms of encryption strength, the method achieves entropy efficiency exceeding 98%, with an average entropy increase of 0.77 bits, representing an improvement of approximately 9.4% over the traditional thumbnail-preserving encryption (TPE) method. Considering the usability of downstream tasks, the proposed method preserves pedestrian detection performance, with F1-scores exceeding 97% in selected scenarios, demonstrating a 0.5% difference compared to TPE while providing substantially stronger privacy protection. The H-FPE method effectively balances privacy protection and functional usability, offering a robust solution for facial data protection in autonomous sensor applications while preserving essential detection capabilities.
Open AccessFace Privacy Protection Method for Autonomous Sensors Based on Hierarchical Format-Preserving Encryption
Advanced sensors in connected automated vehicles (CAVs) increasingly collect facial biometric information for environmental perception, posing serious privacy leakage risks. However, existing privacy protection methods for automotive data primarily focus on strict security mechanisms and fail to fully balance data usability. This paper presents a hierarchical format-preserving encryption (H-FPE) method for face privacy protection in autonomous sensors. The proposed method constructs a privacy-preserving framework for face detection based on YOLOv11 by employing a region-specific encryption strategy where the encryption strength is tailored to the importance of different facial regions. The encryption algorithm employs SM4-based Feistel structures with pseudo-random functions to ensure RGB value constraints while maintaining image format integrity. Experimental evaluation results in diverse scenarios demonstrate that the proposed privacy encryption method achieves superior privacy protection performance. In terms of encryption strength, the method achieves entropy efficiency exceeding 98%, with an average entropy increase of 0.77 bits, representing an improvement of approximately 9.4% over the traditional thumbnail-preserving encryption (TPE) method. Considering the usability of downstream tasks, the proposed method preserves pedestrian detection performance, with F1-scores exceeding 97% in selected scenarios, demonstrating a 0.5% difference compared to TPE while providing substantially stronger privacy protection. The H-FPE method effectively balances privacy protection and functional usability, offering a robust solution for facial data protection in autonomous sensor applications while preserving essential detection capabilities.
- Open Access
- 15 July 2025
- Published by MDPI AG in Electronics
- Vol. 14 (14) , 2843
Face Desensitization for Autonomous Driving Based on Identity De-Identification of Generative Adversarial Networks
Automotive intelligent agents are increasingly collecting facial data for applications such as driver behavior monitoring and identity verification. These excessive collections of facial data bring serious risks of sensitive information leakage to autonomous driving. Facial information has been explicitly required to be anonymized, but the availability of most desensitized facial data is poor, which will greatly affect its application in autonomous driving. This paper proposes an automotive sensitive information anonymization method with high-quality generated facial images by considering the data availability under privacy protection. By comparing K-Same and Generative Adversarial Networks (GANs), this paper proposes a hierarchical self-attention mechanism in StyleGAN3 to enhance the feature perception of face images. The synchronous regularization of sample data is applied to optimize the loss function of the discriminator of StyleGAN3, thereby improving the convergence stability of the model. The experimental results demonstrate that the proposed facial desensitization model reduces the Frechet inception distance (FID) and structural similarity index measure (SSIM) by 95.8% and 24.3%, respectively. The image quality and privacy desensitization of the facial data generated by the StyleGAN3 model have been fully verified in this work. This research provides an efficient and robust facial privacy protection solution for autonomous driving, which is conducive to promoting the security guarantee of automotive data.
- 15 July 2025
- Published by MDPI AG in Electronics
- Vol. 14 (14) , 2843
Open AccessFace Desensitization for Autonomous Driving Based on Identity De-Identification of Generative Adversarial Networks
Automotive intelligent agents are increasingly collecting facial data for applications such as driver behavior monitoring and identity verification. These excessive collections of facial data bring serious risks of sensitive information leakage to autonomous driving. Facial information has been explicitly required to be anonymized, but the availability of most desensitized facial data is poor, which will greatly affect its application in autonomous driving. This paper proposes an automotive sensitive information anonymization method with high-quality generated facial images by considering the data availability under privacy protection. By comparing K-Same and Generative Adversarial Networks (GANs), this paper proposes a hierarchical self-attention mechanism in StyleGAN3 to enhance the feature perception of face images. The synchronous regularization of sample data is applied to optimize the loss function of the discriminator of StyleGAN3, thereby improving the convergence stability of the model. The experimental results demonstrate that the proposed facial desensitization model reduces the Frechet inception distance (FID) and structural similarity index measure (SSIM) by 95.8% and 24.3%, respectively. The image quality and privacy desensitization of the facial data generated by the StyleGAN3 model have been fully verified in this work. This research provides an efficient and robust facial privacy protection solution for autonomous driving, which is conducive to promoting the security guarantee of automotive data.