信息网络安全

信息网络安全 ›› 2026, Vol. 26 ›› Issue (8): 1169-1182.doi: 10.3969/j.issn.1671-1122.2026.08.001

• 学术研究 • 上一篇    下一篇

基于高频特征增强与关键区域保护的人脸伪造检测方法

陈玮1, 田波2, 李顺畅2, 罗光春1, 秦科1(), 王方圆3,4   

  1. 1 电子科技大学计算机科学与工程学院， 成都 611731
    2 云南电信股份有限公司， 昆明 650299
    3 南京理工大学网络空间安全学院， 南京 210094
    4 恒安嘉新（北京）科技股份公司， 北京 100086
  • 收稿日期:2026-01-25 出版日期:2026-08-10 发布日期:2026-09-23
  • 通讯作者: 秦科 E-mail:qinke@uestc.edu.cn
  • 作者简介:陈玮（1980—），男，四川，工程师，博士研究生，主要研究方向为人工智能、网络运营、网信安全和反诈|田波（1981—），男，云南，正高级工程师，本科，主要研究方向为反诈、网络信息安全和人工智能|李顺畅（1979—），男，四川，工程师，本科，主要研究方向为人工智能|罗光春（1974—），男，四川，教授，博士，主要研究方向为云计算与大数据、智能决策|秦科（1980—），男，四川，教授，博士，主要研究方向为人工智能|王方圆（1985—），男，江苏，高级工程师，博士研究生，主要研究方向为大模型网络安全、人工智能
  • 基金资助:
    江苏省前沿技术研发计划(BF2025004)

Face forgery detection method based on high-frequency feature enhancement and key region preservation

Chen Wei1, Tian Bo2, Li Shunchang2, Luo Guangchun1, Qin Ke1(), Wang Fangyuan3,4   

  1. 1 School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China
    2 Yunnan Telecom Co., Ltd., Kunming 650299, China
    3 School of Cyberspace Security, Nanjing University of Science and Technology, Nanjing 210094, China
    4 Heng’an Jiaxin (Beijing) Technology Co., Ltd., Beijing 100086, China
  • Received:2026-01-25 Online:2026-08-10 Published:2026-09-23
  • Contact: Qin Ke E-mail:qinke@uestc.edu.cn

RichHTML

4

PDF (PC)

5

摘要：

随着人脸识别技术在身份认证、金融支付等关键领域的广泛部署，人脸数据的真实性与安全性已成为数据安全治理过程中的核心挑战。现有方法如基于潜空间重建误差依赖扩散模型先验，对仅影响局部区域的细微篡改敏感性不足，局部伪造误差信号易被全局重建误差稀释。因此，文章提出一种基于自编码器的改进重建误差检测方法，区别于扩散模型驱动的潜空间重建误差，该方法融合高频特征增强与关键区域保护双重机制，针对局部篡改检测进行轻量化优化。首先，通过多任务级联卷积神经网络生成空间权重掩码，引导模型关注五官等关键区域；然后，设计轻量化的可学习高通滤波器，自适应地提取并强化图像中的高频伪造痕迹；最后，将二者融合进重建误差计算，实现对误差信号在空间和频域上的协同增强。实验结果表明，该方法在局部篡改检测任务中性能显著提升，局部篡改敏感度达89.4%；在跨数据集测试中，在DeepFakeDetection和WildDeepfake上平均精度分别为92.1%和90.8%，展现出良好的泛化能力；同时，在噪声与压缩扰动下性能下降幅度较小，展现出较强的鲁棒性。

关键词: 数据安全治理, 人脸伪造检测, 潜空间重建误差, 高频特征增强, 关键区域保护

Abstract:

With the widespread deployment of face recognition technology in critical fields such as identity authentication and financial payment, the authenticity and security of face data have become core challenges in data security governance. Existing methods, such as those based on latent space reconstruction error that rely on diffusion model priors, insufficiently sensitive to subtle tampering that only affects local regions, as the local forgery error signals are easily diluted by global reconstruction errors. To address this issue, this paper proposed an improved reconstruction error detection method based on an autoencoder, which was distinct from diffusion-model-driven latent space reconstruction error and integrates dual mechanisms of high-frequency feature enhancement and key region protection, specifically optimized for local tampering detection in a lightweight manner. First, generated spatial weight mask via a multi-task cascaded convolutional network to guide the model to focus on key regions such as facial features; then, it designed a lightweight learnable high-pass filter to adaptively extract and enhance high-frequency forgery traces in images; finally, by incorporating both mechanisms into the reconstruction error calculation, it achieved cooperative enhancement of the error signals in both spatial and frequency domains. Experimental results show that the proposed method achieves significant performance improvement in local tampering detection tasks, with a local tampering sensitivity of 89.4%; in cross-dataset tests, it attains average precision of 92.1% on DeepFakeDetection and 90.8% on WildDeepfake, demonstrating good generalization capability; meanwhile, it exhibits only minor performance degradation under noise and compression perturbations, reflecting strong robustness.

Key words: data security governance, face forgery detection, latent space reconstruction error, high-frequency feature enhancement, key region protection

中图分类号: