Protecting Against Image Translation Deepfakes by Leaking Universal Perturbations from Black-Box Neural Networks.
Nataniel RuizSarah Adel BargalStan SclaroffPublished in: CoRR (2020)
Keyphrases
- black box
- neural network
- input image
- rule extraction
- image data
- image content
- hybrid systems
- image classification
- multiscale
- image features
- image segmentation
- single image
- image analysis
- black boxes
- image representation
- artificial neural networks
- state transition
- integration testing
- segmentation method
- segmentation algorithm
- feature points
- pattern recognition
- pixel values
- edge detection
- high resolution
- databases