Learning Privately from Multiparty Data

Jihun Hamm,Paul Cao,Mikhail Belkin
DOI: https://doi.org/10.48550/arXiv.1602.03552
IF: 5.414
2016-02-10
Machine Learning
Abstract:Learning a classifier from private data collected by multiple parties is an important problem that has many potential applications. How can we build an accurate and differentially private global classifier by combining locally-trained classifiers from different parties, without access to any party's private data? We propose to transfer the `knowledge' of the local classifier ensemble by first creating labeled data from auxiliary unlabeled data, and then train a global $\epsilon$-differentially private classifier. We show that majority voting is too sensitive and therefore propose a new risk weighted by class probabilities estimated from the ensemble. Relative to a non-private solution, our private solution has a generalization error bounded by $O(\epsilon^{-2}M^{-2})$ where $M$ is the number of parties. This allows strong privacy without performance loss when $M$ is large, such as in crowdsensing applications. We demonstrate the performance of our method with realistic tasks of activity recognition, network intrusion detection, and malicious URL detection.
What problem does this paper attempt to address?