7 papers · 1 filter
Class-Conditional Neural Polarizer: A Lightweight and Effective Backdoor Defense by Purifying Poisoned Features
Mingli Zhu, Shaokui Wei, Hongyuan Zha +1
Recent studies have highlighted the vulnerability of deep neural networks to backdoor attacks, where models are manipulated to rely on embedded triggers within poisoned samples, de…
Revisiting the Auxiliary Data in Backdoor Purification
Shaokui Wei, Shanchao Yang, Jiayin Liu +1
Backdoor attacks occur when an attacker subtly manipulates machine learning models during the training phase, leading to unintended behaviors when specific triggers are present. To…
Backdoor Mitigation by Distance-Driven Detoxification
Shaokui Wei, Jiayin Liu, Hongyuan Zha
Backdoor attacks undermine the integrity of machine learning models by allowing attackers to manipulate predictions using poisoned training data. Such attacks lead to targeted misc…
WPDA: Frequency-based Backdoor Attack with Wavelet Packet Decomposition
Zhengyao Song, Yongqiang Li, Danni Yuan +3
This work explores an emerging security threat against deep neural networks (DNNs) based image classification, i.e., backdoor attack. In this scenario, the attacker aims to inject…
Mitigating Backdoor Attack by Injecting Proactive Defensive Backdoor
Shaokui Wei, Hongyuan Zha, Baoyuan Wu
Data-poisoning backdoor attacks are serious security threats to machine learning models, where an adversary can manipulate the training dataset to inject backdoors into models. In…
Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness
Weilin Lin, Li Liu, Shaokui Wei +2
The security threat of backdoor attacks is a central concern for deep neural networks (DNNs). Recently, without poisoned data, unlearning models with clean data and then learning a…