Privacy-Preserving Contrastive Explanations with Local Foil Trees

Thijs Veugen; Bart Kamphorst; Michiel Marcus

Paper 2022/360

Privacy-Preserving Contrastive Explanations with Local Foil Trees

Thijs Veugen, Bart Kamphorst, and Michiel Marcus

Abstract

We present the first algorithm that combines privacy-preserving technologies and state-of-the-art explainable AI to enable privacy-friendly explanations of black-box AI models. We provide a secure algorithm for contrastive explanations of black-box machine learning models that securely trains and uses local foil trees. Our work shows that the quality of these explanations can be upheld whilst ensuring the privacy of both the training data, and the model itself.

Note: To be published in CSCML 2022

Metadata

Available format(s): PDF
Category: Applications
Publication info: Preprint. MINOR revision.
Keywords: Secure multi-party computation explainable AI decision tree
Contact author(s): thijs veugen @ tno nl
History: 2022-03-30: last of 3 revisions; 2022-03-18: received; See all versions
Short URL: https://ia.cr/2022/360
License: CC BY

BibTeX

@misc{cryptoeprint:2022/360,
      author = {Thijs Veugen and Bart Kamphorst and Michiel Marcus},
      title = {Privacy-Preserving Contrastive Explanations with Local Foil Trees},
      howpublished = {Cryptology {ePrint} Archive, Paper 2022/360},
      year = {2022},
      url = {https://eprint.iacr.org/2022/360}
}