A New Backdoor Attack in CNNS by Training Set Corruption Without Label Poisoning.
Mauro BarniKassem KallasBenedetta TondiPublished in: ICIP (2019)
Keyphrases
- training set
- class labels
- data corruption
- test set
- cellular neural networks
- active learning
- training examples
- nearest neighbor
- classification accuracy
- training data
- supervised learning
- countermeasures
- data sets
- multi label
- cross validation
- training samples
- test data
- classification algorithm
- feature space
- base classifiers
- malicious users
- decision trees
- support vector machine
- pairwise
- test images
- active appearance models
- image classification
- face images
- positive and negative examples