Multidomain active defense: Detecting multidomain backdoor poisoned samples via ALL-to-ALL decoupling training without clean datasets.
Binhao MaJiahui WangDejun WangBo MengPublished in: Neural Networks (2023)
Keyphrases
- training dataset
- multi domain
- training set
- training samples
- training examples
- sampling methods
- training data
- input output
- data sets
- number of training samples
- training and testing data
- avoid overfitting
- gene expression datasets
- linear svm
- automatic detection
- training process
- cross domain
- data samples
- benchmark datasets
- supervised learning
- training algorithm
- sample size
- training phase
- intrusion detection
- uci machine learning repository
- semi supervised learning
- domain specific
- learning algorithm
- database