Research map / Adversarial robustness and security / Attacks on text and graphs
LowKey: Leveraging Adversarial Attacks to Protect Social Media Users from Facial Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| protect gallery images with LowKey instead of leaving them clean | FaceScrub identities in Amazon Rekognition · FaceScrub | rank-50 recognition rate: 2.4% | Helped |
| protect gallery images with LowKey instead of using Fawkes | Amazon Rekognition · FaceScrub | recognition rate: 77.5% rank-1; 94.9% rank-50 -> 2.4% rank-50 | Helped |
| protect gallery images with LowKey instead of using Fawkes | Microsoft Azure Face · FaceScrub | recognition rate: more than 74% -> 0.1% | Helped |
| save LowKey-protected images in JPEG instead of PNG | Amazon Rekognition · FaceScrub | recognition rate: 2.4% -> 3.8% | Hurt |
| save LowKey-protected images in JPEG instead of PNG | Microsoft Azure Face · FaceScrub | recognition rate: 0.1% -> 0.2% | Hurt |
| reduce gallery size to five images | LowKey with Amazon Rekognition · FaceScrub, full-size images | rank-50 recognition rate: 32.5% -> 11% | Helped |
| use small probe images | LowKey with Amazon Rekognition · FaceScrub, full-size images | recognition accuracy: 32.5% -> 19% | Helped |
| increase the perceptual similarity penalty from 0.05 to 0.08 to make perturbations smaller (ablation) | LowKey · Amazon Rekognition | rank-50 recognition rate: 17.2% | Hurt |
| increase the perceptual similarity penalty from 0.05 to 0.08 to make perturbations smaller (ablation) | LowKey · Microsoft Azure Face | recognition rate: 5.5% | Hurt |
| add Gaussian smoothing to the attack objective (ablation) | LowKey · facial recognition models tested with defensive Gaussian blur | rank-50 attack effectiveness | Helped |
| apply Gaussian smoothing as a defense before inference (ablation) | facial recognition systems · LowKey-protected images | rank-50 recognition accuracy | No clear effect |
Closest papers by meaning
- Transferable Facial Privacy Protection against Blind Face Restoration via Domain-Consistent Adversarial Obfuscation (ICML 2024 · similarity 0.87)
- Enhancing Facial Privacy Protection via Weakening Diffusion Purification (CVPR 2025 · similarity 0.86)
- Non-Adaptive Adversarial Face Generation (NeurIPS 2025 · similarity 0.85)
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models (AAAI 2026 · similarity 0.85)
- Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval (AAAI 2024 · similarity 0.84)
- DiffAM: Diffusion-based Adversarial Makeup Transfer for Facial Privacy Protection (CVPR 2024 · similarity 0.84)
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.