A prompt-tuned CLIP model trained on five known backdoor attacks can detect a sixth held-out attack type with about 86% accuracy on CIFAR-10 and GTSRB.
Reflection backdoor: A natural backdoor attack on deep neural networks,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Proactive Adversarial Defense: Harnessing Prompt Tuning in Vision-Language Models to Detect Unseen Backdoored Images
A prompt-tuned CLIP model trained on five known backdoor attacks can detect a sixth held-out attack type with about 86% accuracy on CIFAR-10 and GTSRB.