Corpus ID: 53669884

DARCCC: Detecting Adversaries by Reconstruction from Class Conditional Capsules

@article{Frosst2018DARCCCDA,
  title={DARCCC: Detecting Adversaries by Reconstruction from Class Conditional Capsules},
  author={Nicholas Frosst and S. Sabour and Geoffrey E. Hinton},
  journal={ArXiv},
  year={2018},
  volume={abs/1811.06969}
}
We present a simple technique that allows capsule models to detect adversarial images. In addition to being trained to classify images, the capsule model is trained to reconstruct the images from the pose parameters and identity of the correct top-level capsule. Adversarial images do not look like a typical member of the predicted class and they have much larger reconstruction errors when the reconstruction is produced from the top-level capsule for that class. We show that setting a threshold… Expand
27 Citations
Robustifying models against adversarial attacks by Langevin dynamics.
  • 11
  • Highly Influenced
  • PDF
MetaAdvDet: Towards Robust Detection of Evolving Adversarial Attacks
  • 5
  • PDF
DLA: Dense-Layer-Analysis for Adversarial Example Detection
  • 3
  • PDF
MNIST-C: A Robustness Benchmark for Computer Vision
  • 42
  • PDF
Kernelized Capsule Networks
  • 1
  • Highly Influenced
  • PDF
...
1
2
3
...

References

SHOWING 1-10 OF 19 REFERENCES
Adversarial Spheres
  • 151
  • PDF
With Friends Like These, Who Needs Adversaries?
  • 39
  • PDF
Adversarial Manipulation of Deep Representations
  • 180
  • PDF
Adversarial examples in the physical world
  • 2,352
  • PDF
MagNet: A Two-Pronged Defense against Adversarial Examples
  • 566
  • PDF
Explaining and Harnessing Adversarial Examples
  • 6,767
  • PDF
Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methods
  • 940
  • PDF
...
1
2
...