Learning with a Strong Adversary

2015/11/10 by Ruitong Huang, Bing Xu, Huang, Ruitong +6 · 36 citations
Computer Science · #Adversarial Robustness in Machine Learning #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.LG

paper · pdf · doi:10.48550/arxiv.1511.03034

openalex publication_date 2015/11/10 · arxiv created 2016/01/16 · arxiv updated 2016/01/19 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

The robustness of neural networks to intended perturbations has recently attracted significant attention. In this paper, we propose a new method, learning with a strong adversary, that learns robust classifiers from supervised data. The proposed method takes finding adversarial examples as an intermediate step. A new and simple way of finding adversarial examples is presented and experimentally shown to be efficient. Experimental results demonstrate that resulting learning method greatly improves the robustness of the classification models produced.

Citations

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Improving neural networks by preventing co-adaptation of feature detectors
• Gradient-based learning applied to document recognition
• Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups
• MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
• Learning Multiple Layers of Features from Tiny Images

Cited by

• A New Type of Adversarial Examples
• Robustifying Diffusion-Denoised Smoothing Against Covariate Shift
• Using Non-invertible Data Transformations to Build Adversarial-Robust Neural Networks
• Attacks on State-of-the-Art Face Recognition using Attentional Adversarial Attack Generative Network
• Perceptual Deep Neural Networks: Adversarial Robustness through Input Recreation
• Feature Importance Guided Attack: A Model Agnostic Adversarial Attack
• Beneficial Perturbations Network for Defending Adversarial Examples
• Adequacy of the Gradient-Descent Method for Classifier Evasion Attacks
• Linking average- and worst-case perturbation robustness via class selectivity and dimensionality
• Cascade Adversarial Machine Learning Regularized with a Unified Embedding
• Bit Error Robustness for Energy-Efficient DNN Accelerators
• Adversarial Meta-Learning
• Adversarial Examples: Attacks and Defenses for Deep Learning
• Unifying Adversarial Training Algorithms with Flexible Deep Data Gradient Regularization
• GraphDefense: Towards Robust Graph Convolutional Networks
• Improving Robustness of ML Classifiers against Realizable Evasion Attacks Using Conserved Features
• Mitigating Advanced Adversarial Attacks with More Advanced Gradient Obfuscation Techniques
• Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation
• Probably Approximately Correct Constrained Learning
• Adversarial Examples in Modern Machine Learning: A Review
• Fooling a Real Car with Adversarial Traffic Signs
• Playing the Game of Universal Adversarial Perturbations
• Understanding the One-Pixel Attack: Propagation Maps and Locality Analysis
• A Person Re-identification Data Augmentation Method with Adversarial Defense Effect
• Data Quality Matters For Adversarial Training: An Empirical Study
• The Effects of JPEG and JPEG2000 Compression on Attacks using Adversarial Examples
• Adversarial Robustness Assessment: Why both L₀ and L_∞ Attacks Are Necessary
• Subset Scanning Over Neural Network Activations
• Rearchitecting Classification Frameworks For Increased Robustness
• Developing and Defeating Adversarial Examples
• Large Margin Deep Networks for Classification
• ATRO: Adversarial Training with a Rejection Option
• Adversarial Examples Detection in Deep Networks with Convolutional Filter Statistics
• SOAR: Second-Order Adversarial Regularization
• Adversarial Robustness Study of Convolutional Neural Network for Lumbar Disk Shape Reconstruction from MR images
• Practical Convex Formulation of Robust One-hidden-layer Neural Network Training

Related

• Improved Network Robustness with Adversary Critic
• Towards Deep Learning Models Resistant to Adversarial Attacks
• Understanding and Quantifying Adversarial Examples Existence in Linear Classification
• Towards Adversarial Training with Moderate Performance Improvement for Neural Network Classification
• A Survey of Robust Adversarial Training in Pattern Recognition: Fundamental, Theory, and Methodologies