How Deep Learning Sees the World: A Survey on Adversarial Attacks & Defenses

Abstract

Deep Learning is currently used to perform multiple tasks, such as object recognition, face recognition, and natural language processing. However, Deep Neural Networks (DNNs) are vulnerable to perturbations that alter the network prediction (adversarial examples), raising concerns regarding its usage in critical areas, such as self-driving vehicles, malware detection, and healthcare. This paper compiles the most recent adversarial attacks, grouped by the attacker capacity, and modern defenses clustered by protection strategies. We also present the new advances regarding Vision Transformers, summarize the datasets and metrics used in the context of adversarial settings, and compare the state-of-the-art results under different attacks, finishing with the identification of open issues.

References in corpus (35)

• Explaining and Harnessing Adversarial Examples
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Intriguing properties of neural networks
• Improved Regularization of Convolutional Neural Networks with Cutout
• Deep Learning using Rectified Linear Units (ReLU)
• Ensemble Adversarial Training: Attacks and Defenses
• Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks
• DeepXplore: Automated Whitebox Testing of Deep Learning Systems
• Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
• Deep Face Recognition: A Survey
• ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases
• Towards Deep Neural Network Architectures Robust to Adversarial Examples
• DeepGauge: Multi-Granularity Testing Criteria for Deep Learning Systems
• Event-based Vision meets Deep Learning on Steering Prediction for Self-driving Cars
• The Space of Transferable Adversarial Examples
• Stable Architectures for Deep Neural Networks
• Learning with a Strong Adversary
• A study of the effect of JPG compression on adversarial images
• A Dual Approach to Scalable Verification of Deep Networks
• Uncovering the Limits of Adversarial Training against Norm-Bounded Adversarial Examples
• Adversarial Attack Vulnerability of Medical Image Analysis Systems: Unexplored Factors
• Interpreting and Improving Adversarial Robustness of Deep Neural Networks with Neuron Sensitivity
• Fixing Data Augmentation to Improve Adversarial Robustness
• Extending Defensive Distillation
• Evaluating Adversarial Attacks on Driving Safety in Vision-Based Autonomous Vehicles
• Better Diffusion Models Further Improve Adversarial Training
• On the Real-World Adversarial Robustness of Real-Time Semantic Segmentation Models for Autonomous Driving
• On Robustness of Neural Ordinary Differential Equations
• Adversarial Attacks on Machine Learning Systems for High-Frequency Trading
• DensePure: Understanding Diffusion Models towards Adversarial Robustness
• Guided Diffusion Model for Adversarial Purification from Random Noise
• Adversarial Attacks on Remote User Authentication Using Behavioural Mouse Dynamics
• LTD: Low Temperature Distillation for Gradient Masking-free Adversarial Training
• Manifold Regularization for Locally Stable Deep Neural Networks
• Provable Robustness Against a Union of ${\ell }_0$ Adversarial Attacks

Cited by in corpus (3)

• Unleashing the potential of prompt engineering for large language models
• Towards Robust Stability Prediction in Smart Grids: GAN-based Approach under Data Constraints and Adversarial Challenges
• ProARD: progressive adversarial robustness distillation: provide wide range of robust students