CVPR Findings2026
DeepFakeShield: A Proactive Defense Against Malicious Face Swapping
Saeed Karimi-Bidhendi, Joseph DeGol, Eric Wengrowski, Dominic Roberts, Kristin Dana
Abstract
Face-swapping algorithms enable malicious actors to spread misinformation and invade personal privacy and security by inserting victims' faces into images and videos without consent. Existing countermeasures are easily defeated and unfavorably trade perceptual quality with protection efficacy. We present DeepFakeShield (DFS), a proactive method that thwarts face-swapping algorithms by adding adversarial noise that disrupts facial landmark detection. DFS's primary contributions are (1) jointly optimizing both perceptual quality and protection efficacy, and (2) introducing a mutation layer during training to make the shield robust against common removal attacks such as blurring and JPEG compression. Using both qualitative and quantitative results, we show that DFS optimally trades perceptual quality with protection efficacy, remaining robust to image removal attacks and outperforming prior approaches. With ablation studies, we also show that DFS generalizes to out-of-domain face images and that both the joint optimization and mutation layer contribute significantly to our state-of-the-art results.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
Builds on20
- FSGAN: Subject Agnostic Face Swapping and ReenactmentYuval Nirkin, Yosi Keller, Tal HassnerICCV 2019 · 710 citations
- SimSwap: An Efficient Framework For High Fidelity Face SwappingRenwang Chen, Xuanhong Chen, Bingbing Ni, Yanhao GeACM MM 2020 · 409 citations
- Detecting Deepfakes with Self-Blended ImagesKaede Shiohara, Toshihiko YamasakiCVPR 2022 · 366 citations
- ID-Reveal: Identity-aware DeepFake Video DetectionDavide Cozzolino, Andreas Rössler, Justus Thies, Matthias Nießner et al.ICCV 2021 · 216 citations
- Anti-DreamBooth: Protecting users from personalized text-to-image synthesisThanh Van Le, Hao Phung, Thuan Hoang Nguyen, Quan Dao et al.ICCV 2023 · 144 citations
Related papers
- DeepProtect: Proactive Face-Swapping Defense using Identity Blending and Attribute DistortionEungi Lee, Seung-hyeok Back, Hyung-Il Kim, Seok Bong YooCVPR 2026
- FaceShield: Defending Facial Image Against Deepfake ThreatsJaehwan Jeong, Sumin In, Sieun Kim, Hannie Shin et al.ICCV 2025 · 3 citations
- NullSwap: Proactive Identity Cloaking Against Deepfake Face SwappingTianyi Wang, Shuaicheng Niu, Harry Cheng, Xiao Zhang et al.ICCV 2025 · 4 citations
- Phantom: A Unified Face-Swap Deepfake Protection Framework with Latent and Spatial ConstraintsJungkon Kim, Cheolseung Jung, Jong-Min Choi, Juseong LeeCVPR Findings 2026
- PhantomSeal: Proactive Deepfakes Defense with Identity/Context Protection and Forensic TracingLiangqin Ren, Zeyan Liu, Ye Wang, Yuxin Chen et al.CCS 2026
