SemSBA: Semantic-perturbed Stealthy Backdoor Attack on Federated Semi-supervised Learning.
Yingrui TongJun FengGaolei LiXi LinChengcheng ZhaoXiaoyu YiJianhua LiPublished in: ICPADS (2023)
Keyphrases
- semi supervised learning
- semi supervised
- labeled data
- unlabeled data
- supervised learning
- semi supervised classification
- co training
- machine learning
- training data
- label propagation
- unsupervised learning
- improve the classification accuracy
- manifold regularization
- active learning
- labeled examples
- natural language
- semi supervised learning algorithms
- graph based semi supervised learning
- semi supervised learning methods
- labeled and unlabeled data
- learning models
- graph construction
- domain adaptation
- transfer learning
- regularization framework
- data sources
- e learning