Manifold Adversarial

Manifold adversarial research explores how the underlying geometric structure of data (its "manifold") influences the vulnerability of machine learning models to adversarial attacks—small, carefully crafted perturbations that fool the model. Current research focuses on generating adversarial examples that either stay within ("on-manifold") or deviate from ("off-manifold") the data manifold, aiming to improve both the effectiveness of attacks and the robustness of defenses. This work is significant because it provides a deeper understanding of model vulnerabilities and offers avenues for developing more robust and reliable AI systems, particularly in safety-critical applications like facial recognition and autonomous driving.

Papers

May 27, 2024

Exploiting the Layered Intrinsic Dimensionality of Deep Models for Practical Adversarial Training
Enes Altinisik, Safa Messaoud, Husrev Taha Sencar, Hassan Sajjad, Sanjay Chawla
Adversarial Example Adversarial Training Deep Model Better Robustness Intrinsic Dimensionality General Adversarial Manifold Adversarial

February 5, 2024

Transcending Adversarial Perturbations: Manifold-Aided Adversarial Examples with Legitimate Semantics
Shuai Li, Xiaoyu Jiang, Xiaoguang Ma
Adversarial Example Adversarial Perturbation Traditional Adversarial Manifold Adversarial

May 22, 2023

Attribute-Guided Encryption with Facial Texture Masking
Chun Pong Lau, Jiang Liu, Rama Chellappa
Face Image Encryption Method Manifold Adversarial

February 28, 2023

Improving Model Generalization by On-manifold Adversarial Augmentation in the Frequency Domain
Chang Liu, Wenzhao Xiang, Yuan He, Hui Xue, Shibao Zheng, Hang Su
Adversarial Example Model Generalization Frequency Domain Adversarial Augmentation Manifold Augmentation Manifold Adversarial

December 19, 2022

Discrete Point-wise Attack Is Not Enough: Generalized Manifold Adversarial Attack for Face Recognition
Qian Li, Yuxiao Hu, Ye Liu, Dongxiao Zhang, Xin Jin, Yuntian Chen
Adversarial Example Face Recognition Traditional Adversarial Semantic Adversarial Manifold Adversarial

November 21, 2022

Understanding the Vulnerability of Skeleton-based Human Activity Recognition via Black-box Attack
Yunfeng Diao, He Wang, Tianjia Shao, Yong-Liang Yang, Kun Zhou, David Hogg, Meng Wang
Adversarial Attack Black Box Attack Latent Vulnerability Skeleton Based Human Activity Recognition Manifold Adversarial

October 2, 2022

Understanding Adversarial Robustness Against On-manifold Adversarial Examples
Jiancong Xiao, Liusha Yang, Yanbo Fan, Jue Wang, Zhi-Quan Luo
Adversarial Attack Adversarial Example Adversarial Robustness Manifold Adversarial Manifold Value Function

March 24, 2022

A Manifold View of Adversarial Risk
Wenjia Zhang, Yikai Zhang, Xiaoling Hu, Mayank Goswami, Chao Chen, Dimitris Metaxas
Adversarial Risk Flamelet Generated Manifold Manifold Adversarial

Manifold Adversarial

Papers

Exploiting the Layered Intrinsic Dimensionality of Deep Models for Practical Adversarial Training

Transcending Adversarial Perturbations: Manifold-Aided Adversarial Examples with Legitimate Semantics

Attribute-Guided Encryption with Facial Texture Masking

Improving Model Generalization by On-manifold Adversarial Augmentation in the Frequency Domain

Discrete Point-wise Attack Is Not Enough: Generalized Manifold Adversarial Attack for Face Recognition

Understanding the Vulnerability of Skeleton-based Human Activity Recognition via Black-box Attack

Understanding Adversarial Robustness Against On-manifold Adversarial Examples

A Manifold View of Adversarial Risk