Adversarial Instance

Adversarial instances are subtly modified inputs designed to fool machine learning models, primarily deep neural networks (DNNs), into making incorrect predictions. Current research focuses on generating these instances using evolutionary algorithms and adversarial training techniques, analyzing their impact on model robustness and fairness across various architectures including recurrent and diffusion models, and developing methods to detect and mitigate their effects. Understanding and addressing adversarial instances is crucial for improving the reliability and trustworthiness of AI systems in safety-critical applications and advancing the broader field of machine learning.

18papers

Papers

January 19, 2025

Explainable Adversarial Attacks on Coarse-to-Fine Classifiers
Learning Problem Adversarial Instance Adversarial Perturbation Adversarial Attack Traditional Adversarial Layer Wise Relevance Propagation

June 24, 2024

Evaluating the Robustness of Deep-Learning Algorithm-Selection Models by Evolving Adversarial Instances
Adversarial Instance Deep Neural Network Adversarial Sample Native Robustness Deep Recurrent Neural Network Algorithm Selection

May 1, 2024

ASAM: Boosting Segment Anything Model with Adversarial Tuning
Segmentation Task Segment Anything Model Adversarial Fine Tuning Sharpness Aware Minimization Adversarial Instance

April 1, 2024

The Double-Edged Sword of Input Perturbations to Robust Accurate Fairness
Input Perturbation Adversarial Perturbation Robust Fairness Adversarial Bias Adversarial Instance

January 12, 2024

Adversarial Examples are Misaligned in Diffusion Model Manifolds
Adversarial Instance Manifold Diffusion Adversarial Attack Diffusion Model Adversarial Robustness Adversarial Example

November 14, 2023

Towards Improving Robustness Against Common Corruptions in Object Detectors Using Adversarial Contrastive Learning
Adversarial Perturbation Object Detector Adversarial Instance Common Corruption Adversarial Contrastive Learning Adversarial Attack Native Robustness

August 11, 2023

Enhancing Generalization of Universal Adversarial Perturbation through Gradient Aggregation
Enhancing Generalization Aggregated Gradient Adversarial Instance Gradient Aggregation

August 3, 2023

Hard Adversarial Example Mining for Improving Robust Fairness
Robust Fairness Adversarial Training Adversarial Example Adversarial Instance

May 18, 2023

RobustFair: Adversarial Evaluation through Fairness Confusion Directed Gradient Search
Adversarial Evaluation Adversarial Instance Adversarial Perturbation Adversarial Example Native Robustness Fairness Loss

March 13, 2023

Can Adversarial Examples Be Parsed to Reveal Victim Model Information?
Transferable Attack Victim Model Adversarial Instance Adversarial Example Adversarial Attack

December 13, 2022

AdvMIL: Adversarial Multiple Instance Learning for the Survival Analysis on Whole-Slide Images
Whole Slide Image Computational Pathology Adversarial Instance Survival Distribution Whole Slide Survival Analysis Method

August 29, 2022

Reducing Certified Regression to Certified Classification for General Poisoning Attacks
Novel Regression Adversarial Instance Native Robustness Stochastic Defense Model Robustness Poisoning Attack

August 17, 2022

An Evolutionary, Gradient-Free, Query-Efficient, Black-Box Algorithm for Generating Adversarial Instances in Deep Networks
Deep Network Adversarial Data Black Box QuEry Based Attack Adversarial Example Adversarial Instance

June 2, 2022

FACM: Intermediate Layer Still Retain Effective Features against Adversarial Examples
Predictive Distribution Intermediate Layer Effective Feature Adversarial Instance Adversarial Example Adversarial Attack

May 20, 2022

Exploring the Trade-off between Plausibility, Change Intensity and Adversarial Power in Counterfactual Explanations using Multi-objective Optimization
Adversarial Capability Counterfactual Example Adversarial Instance Counterfactual Explanation Explanation Plausibility Multi Objective Optimization Counterfactual Generation

May 2, 2022

MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer
Robust Self Robust Classification Adversarial Robustness Breast Cancer Adversarial Instance Dropout Layer

January 25, 2022

Identifying a Training-Set Attack's Target Using Renormalized Influence Estimation
Adversarial Instance Target Amazon Ally Training Set Attack Influence Estimation

December 14, 2021

On the Impact of Hard Adversarial Instances on Overfitting in Adversarial Training
Adversarial Training Generalization Performance Adversarial Instance Adversarial Learning Adversarial Attack Global Impact

November 13, 2021

Measuring the Contribution of Multiple Model Representations in Detecting Adversarial Instances
Deep Learning Model Adversarial Instance Adversarial Example Adversarial Attack Client Contribution Multiple Model

Adversarial Instance

Papers

Explainable Adversarial Attacks on Coarse-to-Fine Classifiers

Evaluating the Robustness of Deep-Learning Algorithm-Selection Models by Evolving Adversarial Instances

ASAM: Boosting Segment Anything Model with Adversarial Tuning

The Double-Edged Sword of Input Perturbations to Robust Accurate Fairness

Adversarial Examples are Misaligned in Diffusion Model Manifolds

Towards Improving Robustness Against Common Corruptions in Object Detectors Using Adversarial Contrastive Learning

Enhancing Generalization of Universal Adversarial Perturbation through Gradient Aggregation

Hard Adversarial Example Mining for Improving Robust Fairness

RobustFair: Adversarial Evaluation through Fairness Confusion Directed Gradient Search

Can Adversarial Examples Be Parsed to Reveal Victim Model Information?

AdvMIL: Adversarial Multiple Instance Learning for the Survival Analysis on Whole-Slide Images

Reducing Certified Regression to Certified Classification for General Poisoning Attacks

An Evolutionary, Gradient-Free, Query-Efficient, Black-Box Algorithm for Generating Adversarial Instances in Deep Networks

FACM: Intermediate Layer Still Retain Effective Features against Adversarial Examples

Exploring the Trade-off between Plausibility, Change Intensity and Adversarial Power in Counterfactual Explanations using Multi-objective Optimization

MIRST-DM: Multi-Instance RST with Drop-Max Layer for Robust Classification of Breast Cancer

Identifying a Training-Set Attack's Target Using Renormalized Influence Estimation

On the Impact of Hard Adversarial Instances on Overfitting in Adversarial Training

Measuring the Contribution of Multiple Model Representations in Detecting Adversarial Instances