CausAdv: Detecting Adversarial Examples in CNNs Using Causal Reasoning
CausAdv is a novel framework that leverages causal reasoning to detect adversarial examples in Convolutional Neural Networks (CNNs) by analyzing the causal impact of filters on prediction probabilities.
Layer Regression: A Universal and Lightweight Method for Detecting Adversarial Examples
This research paper introduces Layer Regression (LR), a novel defense mechanism against adversarial examples targeting Deep Neural Networks (DNNs) across various domains. LR leverages the inherent sequential architecture of DNNs and the common goal of adversarial attacks to detect manipulated inputs by analyzing changes in layer outputs.
Improving Generalization of Adversarial Example Detection via Principal Adversarial Domain Adaptation
A novel method, Adversarial Example Detection via Principal Adversarial Domain Adaptation (AED-PADA), is proposed to significantly improve the generalization ability of adversarial example detection by identifying Principal Adversarial Domains (PADs) and exploiting multi-source domain adaptation.
© 2024 by Linnk AI