Agnostic Adversarial

Agnostic adversarial research focuses on improving the robustness of machine learning models against adversarial attacks that don't rely on knowledge of the specific model or task. Current efforts explore techniques like optimal transport, certified training methods (including those leveraging diffusion models), and data augmentation strategies to enhance model resilience against these attacks, often employing neural network architectures such as ResNets and Transformers. This work is crucial for building trustworthy AI systems, particularly in safety-critical applications, by mitigating vulnerabilities to sophisticated, model-agnostic attacks that can compromise performance and reliability.

Papers

August 1, 2024

OTAD: An Optimal Transport-Induced Robust Model for Agnostic Adversarial Attack
Kuo Gai, Sicong Wang, Shihua Zhang
Adversarial Training Adversarial Perturbation Optimal Transport Adversarial DEfense Lipschitz Network Optimal Transport Map Agnostic Adversarial

May 15, 2024

Cross-Input Certified Training for Universal Perturbations
Changming Xu, Gagandeep Singh
Trustworthy Machine Learning Agnostic Adversarial Universal Perturbation Injection Based Training

April 10, 2024

Simpler becomes Harder: Do LLMs Exhibit a Coherent Behavior on Simplified Corpora?
Miriam Anschütz, Edoardo Mosca, Georg Groh
Medical LLM Pre Trained Model Text Simplification Rich Input Sentence Simplification Small Corpus Agnostic Adversarial Legible Behavior

April 28, 2023

On the existence of solutions to adversarial training in multiclass classification
Nicolas Garcia Trillos, Matt Jacobs, Jakwang Kim
Adversarial Training Adversarial Perturbation Solution Path Multiclass Classification Robust Classifier Learning Problem Agnostic Adversarial

March 30, 2023

Multiclass Online Learning and Uniform Convergence
Steve Hanneke, Shay Moran, Vinod Raman, Unique Subedi, Ambuj Tewari
Multiclass Classification Uniform Convergence Multiplicative Weight Update Agnostic Adversarial Online Multiclass

July 17, 2022

Threat Model-Agnostic Adversarial Defense using Diffusion Models
Tsachi Blau, Roy Ganz, Bahjat Kawar, Alex Bronstein, Michael Elad
Diffusion Model Adversarial Attack Adversarial Training Adversarial Perturbation Threat Word Imperceptible Attack Agnostic Adversarial

January 17, 2022

Cyberbullying Classifiers are Sensitive to Model-Agnostic Perturbations
Chris Emmery, Ákos Kádár, Grzegorz Chrupała, Walter Daelemans
Toxicity Detection Adversarial Behavior Lexical Substitution Agnostic Adversarial

December 25, 2021

Task and Model Agnostic Adversarial Attack on Graph Neural Networks
Kartik Sharma, Samidha Verma, Sourav Medya, Arnab Bhattacharya, Sayan Ranu
Graph Neural Network Adversarial Attack Adversarial Perturbation Related Task GNN Model Graph Isomorphism Network Agnostic Adversarial

December 16, 2021

All You Need is RAW: Defending Against Adversarial Attacks with Camera Image Pipelines
Yuxuan Zhang, Bo Dong, Felix Heide
Adversarial Attack Adversarial DEfense Adversarial Image Raw Text Agnostic Adversarial Camera Pipeline