DEMYSTIFYING POISONING BACKDOOR ATTACKS FROM A STATISTICAL PERSPECTIVE

Keyphrases

• Poisoning 100%
• Statistical Aspects 100%
• Backdoor Attack 100%
• Benchmark Dataset 14%
• Tight 14%
• Security Threats 14%
• Generative Models 14%
• Upper Boundary 14%
• Lower Boundary 14%
• Diffusion Model 14%
• Learning Model 14%
• Machine Learning Applications 14%
• Malicious Behavior 14%
• Backdoor 14%
• Attack Scenario 14%
• Automatic Trigger 14%
• Discriminative Model 14%
• Large Language Models 14%

Computer Science

• Backdoors 100%
• And-States 12%
• Malicious Behavior 12%
• Generative Model 12%
• Diffusion Model 12%
• Attack Scenario 12%
• Large Language Model 12%
• Machine Learning 12%
• Learning System 12%