Extracting Some Layers of Deep Neural Networks in the Hard-Label Setting

2026
OriginalPaper
Chapter

Published in:

Authors: Isaac A. Canales-Martínez; David Santos
Published in: Progress in Cryptology – LATINCRYPT 2025
Publisher: Springer Nature Switzerland

Read first chapter Log in

Activate our intelligent search to find suitable subject content or patents.

search-config

AI-assisted search

Off

Abstract

This chapter explores the extraction of deep neural network (DNN) parameters in the hard-label setting, where only the most likely class label is provided. It delves into the challenges and recent advancements in efficiently recovering network parameters, including a novel method for extracting the output layer. The text also discusses efficient sign-recovery techniques for contractive-enough layers, providing practical insights for professionals working in the field. The chapter presents experimental results that showcase the effectiveness of these techniques, demonstrating their potential to significantly reduce real execution time. It concludes with a discussion on future research directions, such as developing an end-to-end implementation of the full hard-label attack and studying extraction attacks for networks using different activation functions.

AI Generated

This summary of the content was generated with the help of AI.

Notes

This work was conducted while doing an internship at Technology Innovation Institute, UAE.

Please log in to get access to your license.

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Business + Economics & Engineering + Technology"

Online-Abonnement

Springer Professional "Business + Economics & Engineering + Technology" gives you access to:

more than 102.000 books
more than 537 journals

from the following subject areas:

Automotive
Construction + Real Estate
Business IT + Informatics
Electrical Engineering + Electronics
Energy + Sustainability
Finance + Banking
Management + Leadership
Marketing + Sales
Mechanical Engineering + Materials
Insurance + Risk

Secure your knowledge advantage now!

inform now

Springer Professional "Engineering + Technology"

Online-Abonnement

Springer Professional "Engineering + Technology" gives you access to:

more than 67.000 books
more than 390 journals

from the following specialised fileds:

Automotive
Business IT + Informatics
Construction + Real Estate
Electrical Engineering + Electronics
Energy + Sustainability
Mechanical Engineering + Materials

Secure your knowledge advantage now!

inform now

Springer Professional "Business + Economics"

Online-Abonnement

Springer Professional "Business + Economics" gives you access to:

more than 67.000 books
more than 340 journals

from the following specialised fileds:

Construction + Real Estate
Business IT + Informatics
Finance + Banking
Management + Leadership
Marketing + Sales
Insurance + Risk

Secure your knowledge advantage now!

inform now

previous chapter Generic Partial Decryption as Feature Engineering for Neural Distinguishers

next chapter Rhizomes and the Roots of Efficiency—Improving Prio

Metadata
Footnotes
Literature

Title: Extracting Some Layers of Deep Neural Networks in the Hard-Label Setting
Authors: Isaac A. Canales-Martínez
David Santos
Publisher: Springer Nature Switzerland
Book: Progress in Cryptology – LATINCRYPT 2025
Print ISBN: 978-3-032-06753-1

Electronic ISBN: 978-3-032-06754-8

Copyright Year: 2026
DOI: https://doi.org/10.1007/978-3-032-06754-8_15

This content is only visible if you are logged in and have the appropriate permissions.