FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2967985550> ?p ?o ?g. }

• W2967985550 endingPage "1912" @default.
• W2967985550 startingPage "1895" @default.
• W2967985550 abstract "Differential privacy is a strong notion for privacy that can be used to prove formal guarantees, in terms of a privacy budget, e, about how much information is leaked by a mechanism. When used in privacy-preserving machine learning, the goal is typically to limit what can be inferred from the model about individual training records. However, the calibration of the privacy budget is not well understood. Implementations of privacy-preserving machine learning often select large values of e in order to get acceptable utility of the model, with little understanding of the impact of such choices on meaningful privacy. Moreover, in scenarios where iterative learning procedures are used, relaxed definitions of differential privacy are often used which appear to reduce the needed privacy budget but present poorly understood trade-offs between privacy and utility. In this paper, we quantify the impact of these choices on privacy in experiments with logistic regression and neural network models. Our main finding is that there is no way to obtain privacy for free--relaxed definitions of differential privacy that reduce the amount of noise needed to improve utility also increase the measured privacy leakage. Current mechanisms for differentially private machine learning rarely offer acceptable utility-privacy trade-offs for complex learning tasks: settings that provide limited accuracy loss provide little effective privacy, and settings that provide strong privacy result in useless models." @default.
• W2967985550 created "2019-08-22" @default.
• W2967985550 creator A5026609618 @default.
• W2967985550 creator A5027001546 @default.
• W2967985550 date "2019-08-14" @default.
• W2967985550 modified "2023-10-03" @default.
• W2967985550 title "Evaluating differentially private machine learning in practice" @default.
• W2967985550 cites W112690700 @default.
• W2967985550 cites W1473189865 @default.
• W2967985550 cites W1582126688 @default.
• W2967985550 cites W168548896 @default.
• W2967985550 cites W1985511977 @default.
• W2967985550 cites W1993116423 @default.
• W2967985550 cites W2019735187 @default.
• W2967985550 cites W2022097286 @default.
• W2967985550 cites W2027595342 @default.
• W2967985550 cites W2040228409 @default.
• W2967985550 cites W2051267297 @default.
• W2967985550 cites W2051434435 @default.
• W2967985550 cites W2054514509 @default.
• W2967985550 cites W2086161653 @default.
• W2967985550 cites W2091644709 @default.
• W2967985550 cites W2095133175 @default.
• W2967985550 cites W2099942492 @default.
• W2967985550 cites W2101771965 @default.
• W2967985550 cites W2109135024 @default.
• W2967985550 cites W2109426455 @default.
• W2967985550 cites W2110287632 @default.
• W2967985550 cites W2119874464 @default.
• W2967985550 cites W2146502635 @default.
• W2967985550 cites W2184139426 @default.
• W2967985550 cites W2263253503 @default.
• W2967985550 cites W2294904676 @default.
• W2967985550 cites W2461943168 @default.
• W2967985550 cites W2512472178 @default.
• W2967985550 cites W2520442116 @default.
• W2967985550 cites W2532781556 @default.
• W2967985550 cites W2535690855 @default.
• W2967985550 cites W2591602089 @default.
• W2967985550 cites W2594311007 @default.
• W2967985550 cites W2601012348 @default.
• W2967985550 cites W2605258322 @default.
• W2967985550 cites W2729742878 @default.
• W2967985550 cites W2777914285 @default.
• W2967985550 cites W2778284298 @default.
• W2967985550 cites W2783434283 @default.
• W2967985550 cites W2795435272 @default.
• W2967985550 cites W2798657499 @default.
• W2967985550 cites W2810065831 @default.
• W2967985550 cites W2885216147 @default.
• W2967985550 cites W2887712657 @default.
• W2967985550 cites W2891866930 @default.
• W2967985550 cites W2897830718 @default.
• W2967985550 cites W2902114605 @default.
• W2967985550 cites W2902298865 @default.
• W2967985550 cites W2906133364 @default.
• W2967985550 cites W2932329902 @default.
• W2967985550 cites W2946930197 @default.
• W2967985550 cites W2951857291 @default.
• W2967985550 cites W2953310471 @default.
• W2967985550 cites W2963378725 @default.
• W2967985550 cites W2963385813 @default.
• W2967985550 cites W2963695762 @default.
• W2967985550 cites W2963699739 @default.
• W2967985550 cites W2963828152 @default.
• W2967985550 cites W2963844355 @default.
• W2967985550 cites W2964064173 @default.
• W2967985550 cites W2964121744 @default.
• W2967985550 cites W3118608800 @default.
• W2967985550 cites W6908809 @default.
• W2967985550 cites W92292672 @default.
• W2967985550 hasPublicationYear "2019" @default.
• W2967985550 type Work @default.
• W2967985550 sameAs 2967985550 @default.
• W2967985550 citedByCount "82" @default.
• W2967985550 countsByYear W29679855502018 @default.
• W2967985550 countsByYear W29679855502019 @default.
• W2967985550 countsByYear W29679855502020 @default.
• W2967985550 countsByYear W29679855502021 @default.
• W2967985550 countsByYear W29679855502022 @default.
• W2967985550 crossrefType "proceedings-article" @default.
• W2967985550 hasAuthorship W2967985550A5026609618 @default.
• W2967985550 hasAuthorship W2967985550A5027001546 @default.
• W2967985550 hasConcept C119857082 @default.
• W2967985550 hasConcept C123201435 @default.
• W2967985550 hasConcept C124101348 @default.
• W2967985550 hasConcept C154945302 @default.
• W2967985550 hasConcept C199360897 @default.
• W2967985550 hasConcept C23130292 @default.
• W2967985550 hasConcept C26713055 @default.
• W2967985550 hasConcept C38652104 @default.
• W2967985550 hasConcept C41008148 @default.
• W2967985550 hasConcept C509729295 @default.
• W2967985550 hasConcept C99221444 @default.
• W2967985550 hasConceptScore W2967985550C119857082 @default.
• W2967985550 hasConceptScore W2967985550C123201435 @default.
• W2967985550 hasConceptScore W2967985550C124101348 @default.
• W2967985550 hasConceptScore W2967985550C154945302 @default.

• next