Model Inversion Attacks that Exploit Confidence …

Model Inversion Attacks that Exploit Confidence Informationand Basic CountermeasuresMatt FredriksonCarnegie Mellon UniversitySomesh JhaUniversity of Wisconsin MadisonThomas RistenpartCornell TechABSTRACTM achine-learning (ML) algorithms are increasingly utilizedin privacy-sensitive applications such as predicting lifestylechoices, making medical diagnoses, and facial recognition. Ina Model Inversion attack, recently introduced in a case studyof linear classifiers in personalized medicine by Fredriksonet al. [13], adversarial access to an ML Model is abusedto learn sensitive genomic information about Model Inversion Attacks apply to settings outsidetheirs, however, is develop a new class of Model Inversion attack thatexploits confidence values revealed along with new Attacks are applicable in a variety of settings, andwe explore two in depth: decision trees for lifestyle surveysas used on machine-learning-as-a-service systems and neuralnetworks for facial recognition.

tle concern is that the ability to make prediction queries might enable adversarial clients to back out sensitive data. Recent work by Fredrikson et al. [13] in the context of ge-nomic privacy shows a model inversion attack that is able to use black-box access to prediction models in order to es-timate aspects of someone’s genotype.

Loading..

Tags:

  Model, That, Black, Prediction, Attacks, Inversion, Incom, Exploits, Model inversion attacks that exploit, Of ge nomic

Information