Understanding Model Extraction Games

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The privacy of machine learning models has become a significant concern in many emerging Machine-Learning-as- a-Service applications, where prediction services based on well- trained models are offered to users via the pay-per-query scheme. However, the lack of a defense mechanism can impose a high risk on the privacy of the server's model since an adversary could efficiently steal the model by querying only a few 'good' data points. The game between a server's defense and an adversary's attack inevitably leads to an arms race dilemma, as commonly seen in Adversarial Machine Learning. To study the fundamental tradeoffs between model utility from a benign user's view and privacy from an adversary's view, we develop new metrics to quantify such tradeoffs, analyze their theoretical properties, and develop an optimization problem to understand the optimal adversarial attack and defense strategies. The developed concepts and theory match the empirical findings on the 'equilibrium' between privacy and utility. In terms of optimization, the key ingredient that enables our results is a unified representation of the attack-defense problem as a min-max bi-level problem. The developed results are demonstrated by examples and empirical experiments.

Original languageEnglish (US)
Title of host publicationProceedings - 2022 IEEE 4th International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications, TPS-ISA 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages285-294
Number of pages10
ISBN (Electronic)9781665474085
DOIs
StatePublished - 2022
Event4th IEEE International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications, TPS-ISA 2022 - Virtual, Online, United States
Duration: Dec 14 2022Dec 16 2022

Publication series

NameProceedings - 2022 IEEE 4th International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications, TPS-ISA 2022

Conference

Conference4th IEEE International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications, TPS-ISA 2022
Country/TerritoryUnited States
CityVirtual, Online
Period12/14/2212/16/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

  • Adversarial learning
  • Bi level optimization
  • Machine learning security
  • Minimax optimization

Fingerprint

Dive into the research topics of 'Understanding Model Extraction Games'. Together they form a unique fingerprint.

Cite this