UPSET and ANGRI : Breaking High Performance Image Classifiers
arXiv:1707.01159
Abstract
In this paper, targeted fooling of high performance image classifiers is achieved by developing two novel attack methods. The first method generates universal perturbations for target classes and the second generates image specific perturbations. Extensive experiments are conducted on MNIST and CIFAR10 datasets to provide insights about the proposed algorithms and show their effectiveness.
References in corpus (5)
Cited by in corpus (18)
- Adversarial Example Detection for DNN Models: A Review and Experimental Comparison
- Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- Adversarial Machine Learning in Image Classification: A Survey Towards the Defender's Perspective
- POBA-GA: Perturbation Optimized Black-Box Adversarial Attacks via Genetic Algorithm
- Inspect, Understand, Overcome: A Survey of Practical Methods for AI Safety
- Hermes Attack: Steal DNN Models with Lossless Inference Accuracy
- Towards a Robust and Trustworthy Machine Learning System Development: An Engineering Perspective
- Opportunities and Challenges in Deep Learning Adversarial Robustness: A Survey
- Improving adversarial robustness of deep neural networks by using semantic information
- Security and Privacy for Artificial Intelligence: Opportunities and Challenges
- Heuristic Black-box Adversarial Attacks on Video Recognition Models
- On the Robustness of Human Pose Estimation
- Query-Free Attacks on Industry-Grade Face Recognition Systems under Resource Constraints
- Adversarial Black-Box Attacks On Text Classifiers Using Multi-Objective Genetic Optimization Guided By Deep Networks
- Moving Target Defense for Deep Visual Sensing against Adversarial Examples
- The Vulnerability of the Neural Networks Against Adversarial Examples in Deep Learning Algorithms
- An Evasion Attack against Stacked Capsule Autoencoder
- Adversarial Attacks with Time-Scale Representations