FUBA: Federated uncovering of backdoor attacks for heterogeneous data
Résumé
This paper proposes a post-training defense against pattern-triggered backdoor attacks in federated learning contexts. This approach relies first on the server estimating the attack pattern. The server then provides the estimated pattern to the end-users, who use it directly on their local data to mitigate backdoor attacks during inference time. This scheme offers an improvement over the existing approaches by demonstrating robustness to data heterogeneity among users without needing a shared dataset or additional information from users and regardless of the number of malicious clients. Based on extensive comparison with existing state-of-the-art methods on well-known computer vision datasets, the proposed method is shown to succeed in mitigating backdoor attacks while preserving high accuracy on clean inputs.
Mots clés
artificial intelligence
machine learning
post-training defense
pattern-triggered backdoor attack
online learning
federated learning
attack pattern
defense
end user
inference time
robustness
malicious client
Data privacy
Computer vision
Servers
Security
Intelligent systems
Backdoor Attacks
Attack Patterns
Loss Function
Objective Function
Learning Rate
Classification Task
Global Model
Clean Data
Stochastic Gradient Descent
Target Class
Bandage
Clean Samples
Types Of Attacks
Outlier Values
Local Dataset
Effects Of Attacks
Target Label
MNIST Dataset
Hidden Patterns
Fashion-MNIST
Differential Privacy
Corrupted Data
heterogeneous data
Origine | Fichiers éditeurs autorisés sur une archive ouverte |
---|