New DFCS method enhances backdoor attack efficiency on ML models

Researchers have developed a new method called Distributional Feature Coverage Sample Selection (DFCS) for data-efficient backdoor attacks on machine learning models. This training-free, trigger-agnostic approach clusters fixed pretrained features and selects the nearest sample from each region to convey the trigger-target association. DFCS demonstrated superior performance across various datasets and attack types, achieving an average attack success rate of 96.30% while maintaining clean accuracy. AI

IMPACT This research highlights potential vulnerabilities in machine learning models and could inform the development of more robust defenses against data poisoning attacks.

RANK_REASON The cluster contains an academic paper detailing a new method for backdoor attacks on machine learning models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Blended
• CIFAR-10
• Distributional Feature Coverage Sample Selection
• Imagenette
• Tiny-ImageNet

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →