Property Inference (Senior) on Smile Detection Downstream 10000 samples (train)
0.89AUC (0.1% Ratio)Manipulated Upstream Model (Zero-Activation Attack)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Manipulated Upstream Model (Zero-Activation Attack)Upstream Task=ImageNet Classification, Target Property=Senior2023.03 | 0.89 | 1 | |
| Normal Upstream ModelUpstream Task=ImageNet Classification, Target Property=Senior2023.03 | 0.59 | 0.56 |