Entangled watermarks as a defense against model extraction

Hengrui Jia, Christopher A. Choquette-Choo, Varun Chandrasekaran, Nicolas Papernot

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

Abstract

Machine learning involves expensive data collection and training procedures. Model owners may be concerned that valuable intellectual property can be leaked if adversaries mount model extraction attacks. As it is difficult to defend against model extraction without sacrificing significant prediction accuracy, watermarking instead leverages unused model capacity to have the model overfit to outlier input-output pairs. Such pairs are watermarks, which are not sampled from the task distribution and are only known to the defender. The defender then demonstrates knowledge of the input-output pairs to claim ownership of the model at inference. The effectiveness of watermarks remains limited because they are distinct from the task distribution and can thus be easily removed through compression or other forms of knowledge transfer. We introduce Entangled Watermarking Embeddings (EWE). Our approach encourages the model to learn features for classifying data that is sampled from the task distribution and data that encodes watermarks. An adversary attempting to remove watermarks that are entangled with legitimate data is also forced to sacrifice performance on legitimate data. Experiments on MNIST, Fashion-MNIST, CIFAR-10, and Speech Commands validate that the defender can claim model ownership with 95% confidence with less than 100 queries to the stolen copy, at a modest cost below 0.81 percentage points on average in the defended model's performance.

Original languageEnglish (US)
Title of host publicationProceedings of the 30th USENIX Security Symposium
PublisherUSENIX Association
Pages1937-1954
Number of pages18
ISBN (Electronic)9781939133243
StatePublished - 2021
Externally publishedYes
Event30th USENIX Security Symposium, USENIX Security 2021 - Virtual, Online
Duration: Aug 11 2021Aug 13 2021

Publication series

NameProceedings of the 30th USENIX Security Symposium

Conference

Conference30th USENIX Security Symposium, USENIX Security 2021
CityVirtual, Online
Period8/11/218/13/21

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Safety, Risk, Reliability and Quality

Fingerprint

Dive into the research topics of 'Entangled watermarks as a defense against model extraction'. Together they form a unique fingerprint.

Cite this