Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23453533Reproducibility artifact for the paper Which Face Images Are Easy to Attack? Benchmarking Fragility Estimators Across Face Recognizers (A. Apartsin, Y. Aperstein). The paper benchmarks thirteen per-image adversarial fragility estimators (perturbation probes and face image quality assessment methods, including the ISO/IEC 29794-5 reference implementation OFIQ) against measured white-box, black-box transfer and adaptive attacks on six face recognizers and three public datasets (LFW, VGGFace2, XQLFW).data.zip holds the per-image arrays behind every table: estimator scores, attack start and end similarities at every budget, attacked-probe scores used for quality gating, transfer outcomes, per-dataset face lists (image path, identity label, quality features), the per-recognizer FMR 1e-3 thresholds and the adaptive-attack results. code.zip holds the analysis scripts, the GPU experiment scripts and the result tables.Face images and face embeddings are not included; the datasets are public, and the released scripts re-detect, align and embed the listed faces. Code is released under the MIT license; data and results under CC BY 4.0.
Keyword(s):
adversarial robustness; benchmark; biometrics; face image quality assessment; face recognition.
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)