Abstract
Failure cases of black-box deep learning, e.g. adversarial examples, might have severe consequences in healthcare. Yet such failures are mostly studied in the context of real-world images with calibrated attacks. To demystify the adversarial examples, rigorous studies need to be designed. Unfortunately, complexity of the medical images hinders such study design directly from the medical images. We hypothesize that adversarial examples might result from the incorrect mapping of image space to the low dimensional generation manifold by deep networks. To test the hypothesis, we simplify a complex medical problem namely pose estimation of surgical tools into its barest form. An analytical decision boundary and exhaustive search of the one-pixel attack across multiple image dimensions let us localize the regions of frequent successful one-pixel attacks at the image space.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Berkrot, B.: U.S. FDA approves AI device to detect diabetic eye disease, 11 April 2018. https://www.reuters.com/article/us-fda-ai-approval/u-s-fda-approves-ai-device-to-detect-diabetic-eye-disease-idUSKBN1HI2LC
Eykholt, K., et al.: Robust physical-world attacks on deep learning models (2017). http://arxiv.org/pdf/1707.08945
Finlayson, S.G., Chung, H.W., Kohane, I.S., Beam, A.L.: Adversarial attacks against medical deep learning systems (2018). http://arxiv.org/pdf/1804.05296
Gilmer, J., et al.: Adversarial spheres (2018). http://arxiv.org/pdf/1801.02774
Goodfellow, I.J., Shlens, J., Szegedy, C.: Explaining and harnessing adversarial examples (2014). http://arxiv.org/pdf/1412.6572
Kügler, D., Stefanov, A., Mukhopadhyay, A.: i3PosNet: instrument pose estimation from X-Ray (2018). http://arxiv.org/pdf/1802.09575
Walter, M.: FDA reclassification proposal could ease approval process for CAD software, 01 June 2018. https://www.healthimaging.com/topics/healthcare-economics/fda-reclassification-proposal-could-ease-approval-process-cad-software
Nguyen, A., Yosinski, J., Clune, J.: Deep neural networks are easily fooled: high confidence predictions for unrecognizable images (2014). http://arxiv.org/pdf/1412.1897
Rauber, J., Brendel, W., Bethge, M.: Foolbox: a python toolbox to benchmark the robustness of machine learning models (2017). http://arxiv.org/pdf/1707.04131
Su, J., Vargas, D.V., Kouichi, S.: One pixel attack for fooling deep neural networks (2017). http://arxiv.org/pdf/1710.08864
Szegedy, C., et al.: Intriguing properties of neural networks (2013). http://arxiv.org/pdf/1312.6199
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Kügler, D., Distergoft, A., Kuijper, A., Mukhopadhyay, A. (2018). Exploring Adversarial Examples. In: Stoyanov, D., et al. Understanding and Interpreting Machine Learning in Medical Image Computing Applications. MLCN DLF IMIMIC 2018 2018 2018. Lecture Notes in Computer Science(), vol 11038. Springer, Cham. https://doi.org/10.1007/978-3-030-02628-8_8
Download citation
DOI: https://doi.org/10.1007/978-3-030-02628-8_8
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-02627-1
Online ISBN: 978-3-030-02628-8
eBook Packages: Computer ScienceComputer Science (R0)