信息工程大学学报2026，Vol.27Issue(2)：199-207,9.DOI:10.3969/j.issn.1671-0673.2026.02.010
SAPD:面向遮挡人脸识别的对抗补丁分割防御方法
SAPD:Segmentation-Based Adversarial Patch Defense Framework for Occluded Face Recognition
摘要
Abstract
To address the issues of poor generalization and weak robustness against strong adversarial perturbations in existing face recognition defense methods against adversarial patches,a segmentation-based adversarial patch defense(SAPD)framework is proposed.In this method,a U-Net segmentation network fused with an attention mechanism is constructed,and a multi-scale hybrid loss function is em‑ployed to realize pixel-level patch localization.A morphological mask smoothing(MMS)strategy is in‑troduced to effectively remove perturbations while preserving facial structure.An occluded-area regu‑larization(OAR)loss function is further designed to improve occlusion robustness without adversarial training,maintaining clean-data accuracy.The experiments conducted on the LFW dataset with the added adversarial patches showed that the recognition accuracies of the method for the eyeglass-shaped patches,mask-shaped patch 1,and mask-shaped patch 2 reach 92.50%,96.16%,and 95.67%,respectively.Furthermore,consistent performance gains are observed across mainstream face recogni‑tion models such as MobileFaceNet,FaceNet,and IResNet50,verifying the strong compatibility,gener‑alization capability,and practical value of the proposed approach.关键词
对抗防御/对抗补丁/分割/修复/人脸识别/深度学习/遮挡Key words
adversarial defense/adversarial patch/segmentation/inpainting/face recognition/deep learning/occlude分类
信息技术与安全科学引用本文复制引用
王浩轩,黄瑞阳,刘宏基,汪浣沙..SAPD:面向遮挡人脸识别的对抗补丁分割防御方法[J].信息工程大学学报,2026,27(2):199-207,9.基金项目
军队后勤科研项目(BHQ090003000X03) （BHQ090003000X03）
