A Pragmatic Approach to Membership Inferences on Machine Learning Models
Authors:
- 1 September 2020
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 521-534
Abstract
Membership Inference Attacks (MIAs) aim to determine the presence of a record in a machine learning model's training data by querying the model. Recent work has demonstrated the effectiveness of MIA on various machine learning models and corresponding defenses have been proposed. However, both attacks and defenses have focused on an adversary that indiscriminately attacks all the records without regard to the cost of false positives or negatives. In this work, we revisit membership inference attacks from the perspective of a pragmatic adversary who carefully selects targets and make predictions conservatively. We design a new evaluation methodology that allows us to evaluate the membership privacy risk at the level of individuals and not only in aggregate. We experimentally demonstrate that highly vulnerable records exist even when the aggregate attack precision is close to 50% (baseline). Specifically, on the MNIST dataset, our pragmatic adversary achieves a precision of 95.05% whereas the prior attack only achieves a precision of 51.7%.Keywords
This publication has 21 references indexed in Scilit:
- Privacy-Preserving Deep Learning2015
- ChiNet uncovers rewired transcription subnetworks in tolerant yeast for advanced biofuels conversionNucleic Acids Research, 2015
- Privacy-preserving data exploration in genome-wide association studies2013
- Functional mechanismProceedings of the VLDB Endowment, 2012
- Differential Privacy: On the Trade-Off between Utility and Information Leakage2012
- Protecting Aggregate Genomic DataScience, 2008
- Resolving Individuals Contributing Trace Amounts of DNA to Highly Complex Mixtures Using High-Density SNP Genotyping MicroarraysPLOS Genetics, 2008
- A Fast Learning Algorithm for Deep Belief NetsNeural Computation, 2006
- Combining dependent P-valuesStatistics & Probability Letters, 2002
- Network information criterion-determining the number of hidden units for an artificial neural network modelIEEE Transactions on Neural Networks, 1994