Research map / Adversarial robustness and security / Attacks on text and graphs
FaceShield: Defending Facial Image against Deepfake Threats
Jaehwan Jeong, Sumin In, Sieun Kim, Hannie Shin, Jongheon Jeong, Sang Ho Yoon, Jaewook Chung, Sangpil Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| target cross-attention condition (key/value) path of DM deepfakes instead of the query-image noising path used by AdvDM/Mist/PhotoGuard/SDST | FaceShield · CelebA-HQ, DiffFace | ISM: 0.471 -> 0.243 (vs AdvDM) | Improved |
| add Gaussian blur and low-pass filtering to the noise update | FaceShield · CelebA-HQ | LPIPS: 0.4214 -> 0.2017 (vs AdvDM) | Improved |
| concentrate noise in low frequencies (low-pass filtering) | FaceShield · CelebA-HQ | frequency rate (FR): 2.1077 -> 18.4651 (vs AdvDM) | Improved |
| JPEG compress protected images at quality 75 | FaceShield · CelebA-HQ, IP-Adapter | ISM: 0.072 -> 0.112 | Hurt |
| add ArcFace identity loss L_id (ablation) | FaceShield · CelebA-HQ, DiffFace | ISM: 0.446 -> 0.243 | Improved |
| add face projector loss L_proj (ablation) | FaceShield · CelebA-HQ, SimSwap | ISM: 0.544 -> 0.184 on SimSwap; DiffFace 0.241 -> 0.243 | Mixed |
| add attention disruption loss L_attn (ablation) | FaceShield · CelebA-HQ, DiffFace | ISM: 0.254 -> 0.243 | Improved |
| add MTCNN attack loss L_mtcnn (ablation) | FaceShield · CelebA-HQ, InfoSwap | ISM: 0.354 -> 0.237 on InfoSwap; IP-Adapter 0.047 -> 0.072 | Mixed |
| combine all four losses instead of any single removed (ablation) | FaceShield · six deepfake models, CelebA-HQ | ISM | Improved |
| restrict diffusion attack to UNet mid-layer cross-attention instead of all or up/down layers (ablation) | FaceShield · DM deepfake (qualitative) | visual protection | Improved |
| apply FaceShield to GAN-based deepfake models | SimSwap, InfoSwap · CelebA-HQ | ISM: 0.544 -> 0.184 (SimSwap); 0.431 -> 0.237 (InfoSwap) | Improved |
Closest papers by meaning
- DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion (NeurIPS 2024 · similarity 0.89)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.89)
- Guard Me If You Know Me: Protecting Specific Face-Identity from Deepfakes (NeurIPS 2025 · similarity 0.88)
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.88)
- Where the Devil Hides: Deepfake Detectors Can No Longer Be Trusted (CVPR 2025 · similarity 0.87)
- Adversarial Robust Safeguard for Evading Deep Facial Manipulation (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
