Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning
We consider privacy-preserving learning in the context of online learning. Insettings where data instances arrive sequentially in streaming fashion, incremental trainingalgorithms such as stochastic gradient descent (SGD) can be used to learn and updateprediction models. When labels are costly to ac...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Labor Dynamics Institute
2020-06-01
|
| Series: | The Journal of Privacy and Confidentiality |
| Subjects: | |
| Online Access: | https://journalprivacyconfidentiality.org/index.php/jpc/article/view/720 |
| _version_ | 1811265816604180480 |
|---|---|
| author | Daniel M Bittner Alejandro E Brito Mohsen Ghassemi Shantanu Rane Anand D Sarwate Rebecca N Wright |
| author_facet | Daniel M Bittner Alejandro E Brito Mohsen Ghassemi Shantanu Rane Anand D Sarwate Rebecca N Wright |
| author_sort | Daniel M Bittner |
| collection | DOAJ |
| description | We consider privacy-preserving learning in the context of online learning. Insettings where data instances arrive sequentially in streaming fashion, incremental trainingalgorithms such as stochastic gradient descent (SGD) can be used to learn and updateprediction models. When labels are costly to acquire, active learning methods can beused to select samples to be labeled from a stream of unlabeled data. These labeled datasamples are then used to update the machine learning models. Privacy-preserving onlinelearning can be used to update predictors on data streams containing sensitive information.The differential privacy framework quantifies the privacy risk in such settings. This workproposes a differentially private online active learning algorithm using stochastic gradientdescent (SGD) to retrain the classifiers. We propose two methods for selecting informativesamples. We incorporated this into a general-purpose web application that allows a non-expert user to evaluate the privacy-aware classifier and visualize key privacy-utility tradeoffs.Our application supports linear support vector machines and logistic regression and enablesan analyst to configure and visualize the effect of using differentially private online activelearning versus a non-private counterpart. The application is useful for comparing theprivacy/utility tradeoff of different algorithms, which can be useful to decision makers inchoosing which algorithms and parameters to use. Additionally, we use the application toevaluate our SGD-based solution and to show that it generates predictions with a superiorprivacy-utility tradeoff than earlier methods. |
| first_indexed | 2024-04-12T20:30:45Z |
| format | Article |
| id | doaj.art-8ed838b64a1340a6b69cb186bc6ce0db |
| institution | Directory Open Access Journal |
| issn | 2575-8527 |
| language | English |
| last_indexed | 2024-04-12T20:30:45Z |
| publishDate | 2020-06-01 |
| publisher | Labor Dynamics Institute |
| record_format | Article |
| series | The Journal of Privacy and Confidentiality |
| spelling | doaj.art-8ed838b64a1340a6b69cb186bc6ce0db2022-12-22T03:17:44ZengLabor Dynamics InstituteThe Journal of Privacy and Confidentiality2575-85272020-06-0110210.29012/jpc.720Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active LearningDaniel M Bittner0Alejandro E Brito1Mohsen Ghassemi2Shantanu Rane3Anand D Sarwate4Rebecca N Wright5Rutgers UniversityPalo Alto Research CenterRutgers UniversityPalo Alto Research CenterRutgers UniversityRutgers UniversityWe consider privacy-preserving learning in the context of online learning. Insettings where data instances arrive sequentially in streaming fashion, incremental trainingalgorithms such as stochastic gradient descent (SGD) can be used to learn and updateprediction models. When labels are costly to acquire, active learning methods can beused to select samples to be labeled from a stream of unlabeled data. These labeled datasamples are then used to update the machine learning models. Privacy-preserving onlinelearning can be used to update predictors on data streams containing sensitive information.The differential privacy framework quantifies the privacy risk in such settings. This workproposes a differentially private online active learning algorithm using stochastic gradientdescent (SGD) to retrain the classifiers. We propose two methods for selecting informativesamples. We incorporated this into a general-purpose web application that allows a non-expert user to evaluate the privacy-aware classifier and visualize key privacy-utility tradeoffs.Our application supports linear support vector machines and logistic regression and enablesan analyst to configure and visualize the effect of using differentially private online activelearning versus a non-private counterpart. The application is useful for comparing theprivacy/utility tradeoff of different algorithms, which can be useful to decision makers inchoosing which algorithms and parameters to use. Additionally, we use the application toevaluate our SGD-based solution and to show that it generates predictions with a superiorprivacy-utility tradeoff than earlier methods.https://journalprivacyconfidentiality.org/index.php/jpc/article/view/720Differential PrivacyActive LearningAnomaly Detection |
| spellingShingle | Daniel M Bittner Alejandro E Brito Mohsen Ghassemi Shantanu Rane Anand D Sarwate Rebecca N Wright Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning The Journal of Privacy and Confidentiality Differential Privacy Active Learning Anomaly Detection |
| title | Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning |
| title_full | Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning |
| title_fullStr | Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning |
| title_full_unstemmed | Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning |
| title_short | Understanding Privacy-Utility Tradeoffs in Differentially Private Online Active Learning |
| title_sort | understanding privacy utility tradeoffs in differentially private online active learning |
| topic | Differential Privacy Active Learning Anomaly Detection |
| url | https://journalprivacyconfidentiality.org/index.php/jpc/article/view/720 |
| work_keys_str_mv | AT danielmbittner understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning AT alejandroebrito understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning AT mohsenghassemi understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning AT shantanurane understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning AT ananddsarwate understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning AT rebeccanwright understandingprivacyutilitytradeoffsindifferentiallyprivateonlineactivelearning |