Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Maksym Andriushchenko

Maksym Andriushchenko

Saarland University

On the effectiveness of adversarial training against common corruptions


Mar 03, 2021
Klim Kireev, Maksym Andriushchenko, Nicolas Flammarion

Figure 1 for On the effectiveness of adversarial training against common corruptions
Figure 2 for On the effectiveness of adversarial training against common corruptions
Figure 3 for On the effectiveness of adversarial training against common corruptions
Figure 4 for On the effectiveness of adversarial training against common corruptions

  Access Paper or Ask Questions

RobustBench: a standardized adversarial robustness benchmark


Oct 19, 2020
Francesco Croce, Maksym Andriushchenko, Vikash Sehwag, Nicolas Flammarion, Mung Chiang, Prateek Mittal, Matthias Hein

Figure 1 for RobustBench: a standardized adversarial robustness benchmark
Figure 2 for RobustBench: a standardized adversarial robustness benchmark
Figure 3 for RobustBench: a standardized adversarial robustness benchmark
Figure 4 for RobustBench: a standardized adversarial robustness benchmark

  Access Paper or Ask Questions

Understanding and Improving Fast Adversarial Training


Jul 06, 2020
Maksym Andriushchenko, Nicolas Flammarion

Figure 1 for Understanding and Improving Fast Adversarial Training
Figure 2 for Understanding and Improving Fast Adversarial Training
Figure 3 for Understanding and Improving Fast Adversarial Training
Figure 4 for Understanding and Improving Fast Adversarial Training

  Access Paper or Ask Questions

Sparse-RS: a versatile framework for query-efficient sparse black-box adversarial attacks


Jun 23, 2020
Francesco Croce, Maksym Andriushchenko, Naman D. Singh, Nicolas Flammarion, Matthias Hein

Figure 1 for Sparse-RS: a versatile framework for query-efficient sparse black-box adversarial attacks
Figure 2 for Sparse-RS: a versatile framework for query-efficient sparse black-box adversarial attacks
Figure 3 for Sparse-RS: a versatile framework for query-efficient sparse black-box adversarial attacks
Figure 4 for Sparse-RS: a versatile framework for query-efficient sparse black-box adversarial attacks

  Access Paper or Ask Questions

On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines


Jun 08, 2020
Marius Mosbach, Maksym Andriushchenko, Dietrich Klakow

Figure 1 for On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines
Figure 2 for On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines
Figure 3 for On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines
Figure 4 for On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines

  Access Paper or Ask Questions

Square Attack: a query-efficient black-box adversarial attack via random search


Nov 29, 2019
Maksym Andriushchenko, Francesco Croce, Nicolas Flammarion, Matthias Hein

Figure 1 for Square Attack: a query-efficient black-box adversarial attack via random search
Figure 2 for Square Attack: a query-efficient black-box adversarial attack via random search
Figure 3 for Square Attack: a query-efficient black-box adversarial attack via random search
Figure 4 for Square Attack: a query-efficient black-box adversarial attack via random search

  Access Paper or Ask Questions

Provably Robust Boosted Decision Stumps and Trees against Adversarial Attacks


Jun 08, 2019
Maksym Andriushchenko, Matthias Hein

Figure 1 for Provably Robust Boosted Decision Stumps and Trees against Adversarial Attacks
Figure 2 for Provably Robust Boosted Decision Stumps and Trees against Adversarial Attacks
Figure 3 for Provably Robust Boosted Decision Stumps and Trees against Adversarial Attacks
Figure 4 for Provably Robust Boosted Decision Stumps and Trees against Adversarial Attacks

  Access Paper or Ask Questions

Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem


Dec 13, 2018
Matthias Hein, Maksym Andriushchenko, Julian Bitterwolf

Figure 1 for Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem
Figure 2 for Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem
Figure 3 for Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem
Figure 4 for Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem

  Access Paper or Ask Questions

Logit Pairing Methods Can Fool Gradient-Based Attacks


Oct 29, 2018
Marius Mosbach, Maksym Andriushchenko, Thomas Trost, Matthias Hein, Dietrich Klakow

Figure 1 for Logit Pairing Methods Can Fool Gradient-Based Attacks
Figure 2 for Logit Pairing Methods Can Fool Gradient-Based Attacks
Figure 3 for Logit Pairing Methods Can Fool Gradient-Based Attacks
Figure 4 for Logit Pairing Methods Can Fool Gradient-Based Attacks

  Access Paper or Ask Questions

Provable Robustness of ReLU networks via Maximization of Linear Regions


Oct 17, 2018
Francesco Croce, Maksym Andriushchenko, Matthias Hein

Figure 1 for Provable Robustness of ReLU networks via Maximization of Linear Regions
Figure 2 for Provable Robustness of ReLU networks via Maximization of Linear Regions
Figure 3 for Provable Robustness of ReLU networks via Maximization of Linear Regions
Figure 4 for Provable Robustness of ReLU networks via Maximization of Linear Regions

  Access Paper or Ask Questions

Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation


Nov 05, 2017
Matthias Hein, Maksym Andriushchenko

Figure 1 for Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
Figure 2 for Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
Figure 3 for Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
Figure 4 for Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
* final version accepted at NIPS 2017, fixed bug in implementation of Cross-Lipschitz regularization and lower bound computation, now results are better 
  Access Paper or Ask Questions