Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey (2018-01-02T00:00:00.000000Z)

TL;DR

This paper presents the first comprehensive survey on adversarial attacks on deep learning in computer vision, reviewing the works that design adversarial attack, analyze the existence of such attacks and propose defenses against them.

Abstract

Deep learning is at the heart of the current rise of artificial intelligence. In the field of computer vision, it has become the workhorse for applications ranging from self-driving cars to surveillance and security. Whereas, deep neural networks have demonstrated phenomenal success (often beyond human capabilities) in solving complex problems, recent studies show that they are vulnerable to adversarial attacks in the form of subtle perturbations to inputs that lead a model to predict incorrect outputs. For images, such perturbations are often too small to be perceptible, yet they completely fool the deep learning models. Adversarial attacks pose a serious threat to the success of deep learning in practice. This fact has recently led to a large influx of contributions in this direction. This paper presents the first comprehensive survey on adversarial attacks on deep learning in computer vision. We review the works that design adversarial attacks, analyze the existence of such attacks and propose defenses against them. To emphasize that adversarial attacks are possible in practical conditions, we separately review the contributions that evaluate adversarial attacks in the real-world scenarios. Finally, drawing on the reviewed literature, we provide a broader outlook of this research direction.

Authors

Naveed Akhtar

6 papers

A. Mian

8 papers

TL;DR

Abstract

Authors

References206 items

Robust Physical-World Attacks on Deep Learning Visual Classification

Certified Defenses against Adversarial Examples

NAG: Network for Adversary Generation

Adversary Detection in Neural Networks via Persistent Homology

On the Robustness of Semantic Segmentation Models to Adversarial Attacks

Butterfly Effect: Bidirectional Control of Classification Performance by Small Additive Perturbation

Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing their Input Gradients

Geometric Robustness of Deep Networks: Analysis and Improvement

MagNet and "Efficient Defenses Against Adversarial Attacks" are Not Robust to Adversarial Examples

Adversarial Attacks Beyond the Image Space

Defense Against Universal Adversarial Perturbations

Enhanced Attacks on Defensively Distilled Deep Neural Networks

Intriguing Properties of Adversarial Examples

Mitigating adversarial effects through randomization

HyperNetworks with statistical filtering for defending adversarial examples

Provable defenses against adversarial examples via the convex outer adversarial polytope

Attacking Binarized Neural Networks

PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples

One Pixel Attack for Fooling Deep Neural Networks

Facial Attributes: Accuracy and Adversarial Robustness

Feature-Guided Black-Box Safety Testing of Deep Neural Networks

Mastering the game of Go without human knowledge

Boosting Adversarial Attacks with Momentum

Discovering Adversarial Examples with Momentum

Standard detectors aren't (currently) fooled by physical adversarial stop signs

Soft biometric privacy: Retaining biometric utility of face images while perturbing gender

Ground-Truth Adversarial Examples

Provably Minimally-Distorted Adversarial Examples

Fooling Vision and Language Models Despite Localization and Attention Mechanism

Can you fool AI with adversarial examples on a visual Turing test?

A Learning and Masking Approach to Secure Learning

Art of Singular Vectors and Universal Adversarial Perturbations

Ensemble Methods as a Defense to Adversarial Perturbations Against Deep Neural Networks

Adversarial Examples for Malware Detection

Towards Proving the Adversarial Robustness of Deep Neural Networks

Analysis of classifiers’ robustness to adversarial perturbations

Is Deep Learning Safe for Robot Vision? Adversarial Examples Against the iCub Humanoid

BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain

Towards Interpretable Deep Neural Networks by Leveraging Adversarial Examples

Learning Universal Adversarial Perturbations with Generative Models

Machine Learning as an Adversarial Service: Learning Black-Box Adversarial Examples

ZOO: Zeroth Order Optimization Based Black-box Attacks to Deep Neural Networks without Training Substitute Models

Cascade Adversarial Machine Learning Regularized with a Unified Embedding

Adversarial-Playground: A visualization suite showing how adversarial examples fool deep learning

Robust Physical-World Attacks on Deep Learning Models

Robust Physical-World Attacks on Machine Learning Models

Synthesizing Robust Adversarial Examples

Efficient Defenses Against Adversarial Attacks

Generic Black-Box End-to-End Attack Against State of the Art API Call Based Malware Classifiers

Generic Black-Box End-to-End Attack against RNNs and Other API Calls Based Malware Classifiers

APE-GAN: Adversarial Perturbation Elimination with GAN

Houdini: Fooling Deep Structured Prediction Models

NO Need to Worry about Adversarial Examples in Object Detection in Autonomous Vehicles

Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks

UPSET and ANGRI : Breaking High Performance Image Classifiers

Fast Feature Fool: A data independent approach to universal adversarial perturbations

Towards Deep Learning Models Resistant to Adversarial Attacks

A Closer Look at Memorization in Deep Networks

Adversarial Example Defenses: Ensembles of Weak Defenses are not Strong

Alleviating adversarial attacks via convolutional autoencoder

Feature Squeezing Mitigates and Detects Carlini/Wagner Adversarial Examples

Analysis of universal adversarial perturbations

MagNet: A Two-Pronged Defense against Adversarial Examples

Regularizing deep networks using efficient layerwise adversarial training

Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methods

Ensemble Adversarial Training: Attacks and Defenses

Extending Defensive Distillation

Generative Adversarial Trainer: Defense to Adversarial Perturbations with GAN

Keeping the Bad Guys Out: Protecting and Vaccinating Deep Learning with JPEG Compression

Mitigating fooling with competitive overcomplete output layer neural networks

Parseval Networks: Improving Robustness to Adversarial Examples

Fooling Neural Networks in Face Attractiveness Evaluation: Adversarial Examples with High Attractiveness Score But Low Subjective Score

Universal Adversarial Perturbations Against Semantic Image Segmentation

Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning

The Space of Transferable Adversarial Examples

Enhancing robustness of machine learning systems via data transformations