Norm Attack

Norm attacks, a subfield of adversarial machine learning, aim to evaluate and improve the robustness of deep neural networks (DNNs) by crafting subtly altered inputs designed to cause misclassification. Current research focuses on attacks utilizing various Lp norms, particularly the less-studied L0 norm which prioritizes sparsity of perturbations, and investigates their effectiveness against different DNN architectures, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), as well as multimodal models. Understanding the vulnerabilities revealed by these attacks is crucial for developing more robust and reliable DNNs, particularly in safety-critical applications where adversarial examples pose significant risks. This research also explores methods for certifying robustness against these attacks and for mitigating their impact through techniques like adversarial training and model ensembling.

Papers

August 28, 2024

Evaluating Model Robustness Using Adaptive Sparse L0 Regularization
Weiyou Liu, Zhenyang Li, Weitong Chen
Adversarial Attack Adversarial Example Model Robustness Sparse Regularization Norm Attack

July 25, 2024

Sparse vs Contiguous Adversarial Pixel Perturbations in Multimodal Models: An Empirical Analysis
Cristian-Alexandru Botocan, Raphael Meier, Ljiljana Dolamic
Native Robustness Adversarial Example Many Sparse Multimodal Model Empirical Analysis Norm Attack Unimodal Representation

July 12, 2024

Deep Adversarial Defense Against Multilevel-Lp Attacks
Ren Wang, Yuxuan Li, Alfred Hero
Adversarial Attack Adversarial Robustness Norm Attack

July 6, 2024

A Novel Bifurcation Method for Observation Perturbation Attacks on Reinforcement Learning Agents: Load Altering Attacks on a Cyber Physical Power System
Kiernan Broda-Milian, Ranwa Al-Mallah, Hanane Dagdougui
Adversarial Attack Deep Reinforcement Learning Adversarial Example New Attack Reinforcement Learning Agent Renewable Energy Cyber Physical Bifurcation Diagram Untargeted Attack Norm Attack

June 4, 2024

Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing
Youwei Shu, Xi Xiao, Derui Wang, Yuxin Cao, Siji Chen, Jason Xue, Linyi Li, Bo Li
Mixed Effect Randomized Smoothing Gaussian Distribution Robustness Certification Smoothing Factor Norm Attack

February 9, 2024

RAMP: Boosting Adversarial Robustness Against Multiple $l_p$ Perturbations for Universal Robustness
Enyi Jiang, Gagandeep Singh
Adversarial Attack Adversarial Training Adversarial Robustness \Ell_p$ Norm Norm Attack Robustness Trade

February 2, 2024

$σ$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples
Antonio Emanuele Cinà, Francesco Villani, Maura Pintor, Lea Schönherr, Battista Biggio, Marcello Pelillo
Adversarial Example Optimization Purpose Adversarial Robustness Norm Attack Sparse Attack \Ell_1$ Norm

October 13, 2023

Is Certifying $\ell_p$ Robustness Still Worthwhile?
Ravi Mangal, Klas Leino, Zifan Wang, Kai Hu, Weicheng Yu, Corina Pasareanu, Anupam Datta, Matt Fredrikson
Adversarial Attack Threat Model Robustness Certification Certified Robustness Worth Multiple Word Norm Attack \Ell_p$ Robustness

August 12, 2023

Not So Robust After All: Evaluating the Robustness of Deep Neural Networks to Unseen Adversarial Attacks
Roman Garaev, Bader Rasheed, Adil Khan
Deep Neural Network Adversarial Attack Native Robustness Norm Attack

March 17, 2023

Robust Mode Connectivity-Oriented Adversarial Defense: Enhancing Neural Network Robustness Against Diversified $\ell_p$ Attacks
Ren Wang, Yuxuan Li, Sijia Liu
Adversarial Attack Adversarial Robustness Adversarial DEfense Neural Network Robustness Norm Attack \Ell_p$ Robustness

February 20, 2023

Seasoning Model Soups for Robustness to Adversarial and Natural Distribution Shifts
Francesco Croce, Sylvestre-Alvise Rebuffi, Evan Shelhamer, Sven Gowal
Native Robustness Adversarial Training \Ell_p$ Norm Model Soup Norm Attack Natural Distribution Shift

October 5, 2022

A Closer Look at Robustness to L-infinity and Spatial Perturbations and their Composition
Luke Rowe, Benjamin Thérien, Krzysztof Czarnecki, Hongyang Zhang
Native Robustness Adversarial Learning Glance Annotation Compositional Ability Norm Attack Composite Attack Compositional Adversarial

March 9, 2022

Reverse Engineering $\ell_p$ attacks: A block-sparse optimization approach with recovery guarantees
Darshan Thaker, Paris Giampouras, René Vidal
Adversarial Attack Reverse Engineering Attack Method Recovery Guarantee \Ell_p$ Norm Imperceptible Perturbation Norm Attack

Norm Attack

Papers

Evaluating Model Robustness Using Adaptive Sparse L0 Regularization

Sparse vs Contiguous Adversarial Pixel Perturbations in Multimodal Models: An Empirical Analysis

Deep Adversarial Defense Against Multilevel-Lp Attacks

A Novel Bifurcation Method for Observation Perturbation Attacks on Reinforcement Learning Agents: Load Altering Attacks on a Cyber Physical Power System

Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing

RAMP: Boosting Adversarial Robustness Against Multiple $l_p$ Perturbations for Universal Robustness

$σ$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples

Is Certifying $\ell_p$ Robustness Still Worthwhile?

Not So Robust After All: Evaluating the Robustness of Deep Neural Networks to Unseen Adversarial Attacks

Robust Mode Connectivity-Oriented Adversarial Defense: Enhancing Neural Network Robustness Against Diversified $\ell_p$ Attacks

Seasoning Model Soups for Robustness to Adversarial and Natural Distribution Shifts

A Closer Look at Robustness to L-infinity and Spatial Perturbations and their Composition

Reverse Engineering $\ell_p$ attacks: A block-sparse optimization approach with recovery guarantees