The paper "Privacy-Preserving Face Recognition Using Trainable Feature Subtraction" addresses the growing concern of privacy in face recognition systems, where unauthorized access to face images can expose sensitive personal information. The authors propose a novel method called MinusFace to protect face images from viewing and recovery attacks. Inspired by image compression techniques, MinusFace creates a visually uninformative face image by subtracting the original face from its model-produced regeneration. Recognizable identity features are preserved by co-training a recognition model on the high-dimensional feature representation. To enhance privacy, the high-dimensional representation is crafted through random channel shuffling, resulting in randomized recognizable images devoid of attacker-leverageable texture details. The methodology is implemented as a privacy-preserving face recognition framework, MinusFace, which demonstrates high recognition accuracy and effective privacy protection. Experimental results show that MinusFace outperforms state-of-the-art (SOTA) methods in both recognition accuracy and privacy protection. The code for MinusFace is available at <https://github.com/Tencent/TFace>.The paper "Privacy-Preserving Face Recognition Using Trainable Feature Subtraction" addresses the growing concern of privacy in face recognition systems, where unauthorized access to face images can expose sensitive personal information. The authors propose a novel method called MinusFace to protect face images from viewing and recovery attacks. Inspired by image compression techniques, MinusFace creates a visually uninformative face image by subtracting the original face from its model-produced regeneration. Recognizable identity features are preserved by co-training a recognition model on the high-dimensional feature representation. To enhance privacy, the high-dimensional representation is crafted through random channel shuffling, resulting in randomized recognizable images devoid of attacker-leverageable texture details. The methodology is implemented as a privacy-preserving face recognition framework, MinusFace, which demonstrates high recognition accuracy and effective privacy protection. Experimental results show that MinusFace outperforms state-of-the-art (SOTA) methods in both recognition accuracy and privacy protection. The code for MinusFace is available at <https://github.com/Tencent/TFace>.