Natural Adversarial

Natural adversarial research focuses on improving the robustness of machine learning models, particularly deep neural networks, against unexpected or "natural" inputs that cause failures, unlike traditional adversarial attacks. Current research explores methods for generating and quantifying these natural adversarial examples, often leveraging techniques like low-frequency perturbations or copy-paste attacks, and analyzing the resulting "natural-adversarial frontier" to understand model vulnerabilities. This work is crucial for enhancing the reliability and safety of AI systems in real-world applications, especially in human-robot interaction and autonomous driving, where unexpected inputs are common.

Papers

February 4, 2024

Spin: An Efficient Secure Computation Framework with GPU Acceleration
Wuxuan Jiang, Xiangjun Song, Shenbai Hong, Haijun Zhang, Wenxin Liu, Bo Zhao, Wei Xu, Yi Li
Convolutional Neural Network Attention Mechanism Training Deep Secure Computation Spin Direction Natural Adversarial

October 31, 2023

LFAA: Crafting Transferable Targeted Adversarial Examples with Low-Frequency Perturbations
Kunyu Wang, Juluan Shi, Wenxuan Wang
Adversarial Attack Adversarial Example Adversarial Perturbation High Frequency Adversarial Color Natural Adversarial

October 16, 2023

Quantifying Assistive Robustness Via the Natural-Adversarial Frontier
Jerry Zhi-Yang He, Zackory Erickson, Daniel S. Brown, Anca D. Dragan
Native Robustness Robust Policy Adversarial Behavior Adversarial Policy Natural Adversarial Collaborative Reinforcement Learning

September 29, 2023

Intrinsic Biologically Plausible Adversarial Robustness
Matilde Tristany Farinha, Thomas Ortner, Giorgia Dellaferrera, Benjamin Grewe, Angeliki Pantazi
Adversarial Attack Adversarial Example Adversarial Robustness Adversarial Sample Natural Adversarial

September 27, 2023

The Robust Semantic Segmentation UNCV2023 Challenge Results
Xuanlong Yu, Yi Zuo, Zitao Wang, Xiaowen Zhang, Jiaxuan Zhao, Yuting Yang, Licheng Jiao, Rui Peng, Xinyi Wang, Junpei Zhang, Kexin Zhang, Fang Liu, Roberto Alcover-Couso, Juan C. SanMiguel, Marcos Escudero-Viñolo, Hanlin Tian, Kenta Matsui, Tianhao Wang, Fahmy Adan, Zhitong Gao, Xuming He, Quentin Bouniot, Hossein Moghaddam, Shyam Nandan Rai, Fabio Cermelli, Carlo Masone, Andrea Pilzer, Elisa Ricci, Andrei Bursuc, Arno Solin, Martin Trapp, Rui Li, Angela Yao, Wenlong Chen, Ivor Simpson, Neill D. F. Campbell, Gianni Franchi
Semantic Segmentation Computer Vision Uncertainty Quantification Robust Semantic Segmentation Natural Adversarial

January 27, 2023

Alignment with human representations supports robust few-shot learning
Ilia Sucholutsky, Thomas L. Griffiths
Shot Learning Alignment Problem Computer Vision Model Representation Alignment Human Representation Aligned Model Natural Adversarial ROBust Few Shot

November 18, 2022

Diagnostics for Deep Neural Networks with Automated Copy/Paste Attacks
Stephen Casper, Kaivalya Hariharan, Dylan Hadfield-Menell
Deep Neural Network Adversarial Example Disease Diagnostics Natural Adversarial Paste Attack