Connexion
Idée - Backdoor Attack Defense
PAD-FT: A Lightweight Defense Against Backdoor Attacks via Data Purification and Fine-Tuning
A lightweight defense mechanism, PAD-FT, that effectively disinfects poisoned deep neural network models without requiring additional clean data.
Enhancing Backdoor Attack Survivability with Venom
Venom proposes a novel approach to enhance the survivability of backdoor attacks against model reconstruction-based defenses by coupling decision paths, preserving attack capabilities while improving survivability.
Mitigating Backdoor Attacks on Language Models through Model Merging
The author proposes using model merging as an effective defense against backdoor attacks on language models, showcasing robustness and versatility in various contexts.
Produits
© 2024 by Linnk AI