An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning

Kumar, Mohit; Moser, Bernhard A.; Fischer, Lukas; Freudenthaler, Bernhard

doi:10.3390/a16090450

Open AccessArticle

An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning

by

Mohit Kumar

^1,2,*

,

Bernhard A. Moser

^2,3,

Lukas Fischer

²

and

Bernhard Freudenthaler

²

¹

Faculty of Computer Science and Electrical Engineering, University of Rostock, 18051 Rostock, Germany

²

Software Competence Center Hagenberg GmbH, A-4232 Hagenberg, Austria

³

Institute of Signal Processing, Johannes Kepler University Linz, 4040 Linz, Austria

^*

Author to whom correspondence should be addressed.

Algorithms 2023, 16(9), 450; https://doi.org/10.3390/a16090450

Submission received: 7 June 2023 / Revised: 30 August 2023 / Accepted: 8 September 2023 / Published: 20 September 2023

(This article belongs to the Special Issue Deep Learning Techniques for Computer Security Problems)

Download

Browse Figures

Versions Notes

Abstract

In order to develop machine learning and deep learning models that take into account the guidelines and principles of trustworthy AI, a novel information theoretic approach is introduced in this article. A unified approach to privacy-preserving interpretable and transferable learning is considered for studying and optimizing the trade-offs between the privacy, interpretability, and transferability aspects of trustworthy AI. A variational membership-mapping Bayesian model is used for the analytical approximation of the defined information theoretic measures for privacy leakage, interpretability, and transferability. The approach consists of approximating the information theoretic measures by maximizing a lower-bound using variational optimization. The approach is demonstrated through numerous experiments on benchmark datasets and a real-world biomedical application concerned with the detection of mental stress in individuals using heart rate variability analysis.

Keywords: privacy; interpretability; transferability; information theory; membership mappings; variational optimization; machine and deep learning

Share and Cite

MDPI and ACS Style

Kumar, M.; Moser, B.A.; Fischer, L.; Freudenthaler, B. An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning. Algorithms 2023, 16, 450. https://doi.org/10.3390/a16090450

AMA Style

Kumar M, Moser BA, Fischer L, Freudenthaler B. An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning. Algorithms. 2023; 16(9):450. https://doi.org/10.3390/a16090450

Chicago/Turabian Style

Kumar, Mohit, Bernhard A. Moser, Lukas Fischer, and Bernhard Freudenthaler. 2023. "An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning" Algorithms 16, no. 9: 450. https://doi.org/10.3390/a16090450

APA Style

Kumar, M., Moser, B. A., Fischer, L., & Freudenthaler, B. (2023). An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning. Algorithms, 16(9), 450. https://doi.org/10.3390/a16090450

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

An Information Theoretic Approach to Privacy-Preserving Interpretable and Transferable Learning

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI