Differentially private online active learning with applications to anomaly detection

Mohsen Ghassemi, Anand D. Sarwate, Rebecca N. Wright

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

6 Scopus citations

Abstract

In settings where data instances are generated sequentially or in streaming fashion, online learning algorithms can learn predictors using incremental training algorithms such as stochastic gradient descent. In some security applications such as training anomaly detectors, the data streams may consist of private information or transactions and the output of the learning algorithms may reveal information about the training data. Differential privacy is a framework for quantifying the privacy risk in such settings. This paper proposes two differentially private strategies to mitigate privacy risk when training a classifier for anomaly detection in an online setting. The first is to use a randomized active learning heuristic to screen out uninformative data points in the stream. The second is to use mini-batching to improve classifier performance. Experimental results show how these two strategies can trade off privacy, label complexity, and generalization performance.

Original languageEnglish (US)
Title of host publicationAISec 2016 - Proceedings of the 2016 ACM Workshop on Artificial Intelligence and Security, co-located with CCS 2016
PublisherAssociation for Computing Machinery, Inc
Pages117-128
Number of pages12
ISBN (Electronic)9781450345736
DOIs
StatePublished - Oct 28 2016
Event9th ACM Workshop on Artificial Intelligence and Security, AISec 2016 - Vienna, Austria
Duration: Oct 28 2016 → …

Publication series

NameAISec 2016 - Proceedings of the 2016 ACM Workshop on Artificial Intelligence and Security, co-located with CCS 2016

Other

Other9th ACM Workshop on Artificial Intelligence and Security, AISec 2016
Country/TerritoryAustria
CityVienna
Period10/28/16 → …

ASJC Scopus subject areas

  • Artificial Intelligence

Keywords

  • Active learning
  • Anomaly detection
  • Differential privacy
  • Online learning
  • Stochastic gradient descent

Fingerprint

Dive into the research topics of 'Differentially private online active learning with applications to anomaly detection'. Together they form a unique fingerprint.

Cite this