A little is enough: Circumventing defenses for distributed learning

Keyphrases

• Defense Mechanisms 100%
• Distributed Learning 100%
• MNIST 100%
• CIFAR-100 100%
• Byzantine Empire 100%
• Attacker 50%
• Attack Strategy 50%
• Deep Learning 50%
• Learning Process 50%
• Behavioral Model 50%
• Security Threats 50%
• Model Accuracy 50%
• Repurposing 50%
• Hurt 50%
• Attack Model 50%
• Population Mean 50%
• Backdoor 50%
• Rogue 50%
• CIFAR-10 50%
• Population Variance 50%
• Empirical Variance 50%
• Backdooring 50%
• Large-scale Training 50%
• CIFAR-10 Dataset 50%
• MNIST Dataset 50%

Computer Science

• Distributed Learning 100%
• Attackers 50%
• Behavior Model 50%
• Deep Learning Model 50%
• Learning Process 50%
• Attack Model 50%
• Backdoors 50%