Stealing Machine Learning Models via Prediction APIs

Stealing Machine Learning Models via Prediction APIs Florian Tram r, cole Polytechnique F d rale de Lausanne (EPFL); Fan Zhang, Cornell University; Ari Juels, Cornell Tech; Michael K. Reiter, The University of North Carolina at Chapel Hill; Thomas Ristenpart, Cornell Tech This paper is included in the Proceedings of the 25th USENIX Security Symposium August 10 12, 2016 Austin, TX. ISBN 978-1-931971-32-4. Open access to the Proceedings of the 25th USENIX Security Symposium is sponsored by USENIX. Stealing Machine Learning Models via Prediction APIs Florian Trame r Fan Zhang Ari Juels EPFL Cornell University Cornell Tech, Jacobs Institute Michael K. Reiter Thomas Ristenpart UNC Chapel Hill Cornell Tech Abstract that can generate labels on future inputs. Popular Models Machine Learning (ML) Models may be deemed con- include support vector machines (SVMs), logistic regres- fidential due to their sensitive training data, commercial sions, neural networks, and decision trees.

Stealing Machine Learning Models via Prediction APIs ... In such attacks, an adversary with black-box ac-cess, but no prior knowledge of an ML model’s param- ... In summary, we explore model extraction attacks, a practical kind of learning task that, in particular, affects

Loading..

Tags:

  Practical, Machine, Learning, Black, Attacks, Machine learning

Information