Cleverhans V0.1: an Adversarial Machine Learning Library

By Ian J. Goodfellow, Nicolas Papernot, Patrick Drew McDaniel. 2016

314 citations recorded by the source corpus.

Subfields: Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Safety, Robustness & Interpretability

Citations in the corpus

Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, positive): “For the particular case of DL models, available solutions such as Cleverhans [360] seek to detect adversarial vulnerabilities, and provide diﬀerent approaches to harden the model against them.”
Evidential Deep Learning to Quantify Classification Uncertainty (2018, methodology): “For each model trained in the previous experiments, adversarial examples are generated using the Fast Gradient Sign method from the Cleverhans adversarial machine learning library [28], using various values of adversarial perturbation coefﬁcient (cid:15) .”
Spectral Signatures in Backdoor Attacks (2018, background): “We also point out that backdoor poisoning is related to adversarial examples [13, 26, 21, 12, 29, 4, 24, 32].”
Backdooring Convolutional Neural Networks via Targeted Weight Perturbations (2018, background): “present but performed at state-of-the-art accuracy in all other cases. The other, more prevalent class of attacks leveraging perturbations that has been studied is referred to as adversarial examples [10, 19, 7, 31, 21, 23, 25, 29]. This class of attacks targets the images being classiﬁed. It has be”
Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks (2017, methodology): “To evaluate the effectiveness of composing feature squeezing with adversarial training, we combined it with two different adversarial training methods: the FGSM-based version implemented in Cleverhans [32] and the PGD-based version implemented by Madry et al.”
On the (Statistical) Detection of Adversarial Examples (2017, background): “1 library [12] and varied the perturbation in the experiments.”
Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models (2018, methodology): “Defensive distillation (Papernot et al., 2016d) trains the classiﬁer in two rounds using a variant of the distillation (Hinton et al., 2014) method.”
Certified Defenses against Adversarial Examples (2018, background): “, 2017); many of these attacks are collated in the cleverhans repository (Goodfellow et al., 2016).”
Mitigating adversarial effects through randomization (2017, methodology): “Recent works (Papernot et al., 2016b; Kurakin et al., 2017; Tram`er et al., 2017; Metzen et al., 2017; Feinman et al., 2017; Meng & Chen, 2017) are making progress on this line of research.”
Robust Physical-World Attacks on Deep Learning Models (2017, methodology): “The network we used was originally defined in the Cleverhans library [20] and consists of three convolutional layers followed by a fully connected layer (see Table II).”
MagNet: A Two-Pronged Defense against Adversarial Examples (2017, methodology): “In this setting, we evaluated MagNet on popular attacks [26, 22, 2].”
Safety Verification of Deep Neural Networks (2016, methodology): “Similar (statistical) robustness properties of deep neural network classiﬁers, which compute the average minimum distance to a misclassiﬁcation and are independent of the data point, have been studied and can be estimated using tools such as DeepFool [25] and cleverhans [27].”
Adversarial Examples: Attacks and Defenses for Deep Learning (2017, background): “Cleverhans [150] and Foolbox [151] are open-source libraries to benchmark the vulnerability of deep neural networks against adversarial images.”
Dropout Inference in Bayesian Neural Networks with Alpha-divergences (2017, methodology): “We use the single gradient step FGS implemented in Cleverhans (Papernot et al., 2016) with the stepsize η varied”
Multiplicative Normalizing Flows for Variational Bayesian Neural Networks (2017, methodology): “We also measure how robust our models and uncertainties are against adversarial examples (Szegedy et al., 2013; Goodfellow et al., 2014) by generating examples using the fast sign method (Goodfellow et al., 2014) for each of the previously trained architectures using Cleverhans (Paper-not et al., 20”
Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks (2017, methodology): “The attack is implemented using the Cleverhans library [46].”
Detecting Adversarial Samples from Artifacts (2017, methodology): “Jacobian-based Saliency Map Attack (JSMA): Papernot et al. (2016b) proposed a simple iterative method for targeted misclassification.”