Selective Feature Anonymization for Privacy-Preserving Image Data Publishing
There is a strong positive correlation between the development of deep learning and the amount of public data available. Not all data can be released in their raw form because of the risk to the privacy of the related individuals. The main objective of privacy-preserving data publication is to anony...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-05-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/9/5/874 |
| _version_ | 1797567076084219904 |
|---|---|
| author | Taehoon Kim Jihoon Yang |
| author_facet | Taehoon Kim Jihoon Yang |
| author_sort | Taehoon Kim |
| collection | DOAJ |
| description | There is a strong positive correlation between the development of deep learning and the amount of public data available. Not all data can be released in their raw form because of the risk to the privacy of the related individuals. The main objective of privacy-preserving data publication is to anonymize the data while maintaining their utility. In this paper, we propose a privacy-preserving semi-generative adversarial network (PPSGAN) that selectively adds noise to class-independent features of each image to enable the processed image to maintain its original class label. Our experiments on training classifiers with synthetic datasets anonymized with various methods confirm that PPSGAN shows better utility than other conventional methods, including blurring, noise-adding, filtering, and generation using GANs. |
| first_indexed | 2024-03-10T19:36:28Z |
| format | Article |
| id | doaj.art-441b3d5711374e7687a2a8f3fedb9cc9 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-10T19:36:28Z |
| publishDate | 2020-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-441b3d5711374e7687a2a8f3fedb9cc92023-11-20T01:36:09ZengMDPI AGElectronics2079-92922020-05-019587410.3390/electronics9050874Selective Feature Anonymization for Privacy-Preserving Image Data PublishingTaehoon Kim0Jihoon Yang1Machine Learning Research Laboratory, Department of Computer Science and Engineering, Sogang University, Seoul 04107, KoreaMachine Learning Research Laboratory, Department of Computer Science and Engineering, Sogang University, Seoul 04107, KoreaThere is a strong positive correlation between the development of deep learning and the amount of public data available. Not all data can be released in their raw form because of the risk to the privacy of the related individuals. The main objective of privacy-preserving data publication is to anonymize the data while maintaining their utility. In this paper, we propose a privacy-preserving semi-generative adversarial network (PPSGAN) that selectively adds noise to class-independent features of each image to enable the processed image to maintain its original class label. Our experiments on training classifiers with synthetic datasets anonymized with various methods confirm that PPSGAN shows better utility than other conventional methods, including blurring, noise-adding, filtering, and generation using GANs.https://www.mdpi.com/2079-9292/9/5/874adversarial learningdata privacydeep learningdifferential privacygenerative adversarial networksmachine learning |
| spellingShingle | Taehoon Kim Jihoon Yang Selective Feature Anonymization for Privacy-Preserving Image Data Publishing Electronics adversarial learning data privacy deep learning differential privacy generative adversarial networks machine learning |
| title | Selective Feature Anonymization for Privacy-Preserving Image Data Publishing |
| title_full | Selective Feature Anonymization for Privacy-Preserving Image Data Publishing |
| title_fullStr | Selective Feature Anonymization for Privacy-Preserving Image Data Publishing |
| title_full_unstemmed | Selective Feature Anonymization for Privacy-Preserving Image Data Publishing |
| title_short | Selective Feature Anonymization for Privacy-Preserving Image Data Publishing |
| title_sort | selective feature anonymization for privacy preserving image data publishing |
| topic | adversarial learning data privacy deep learning differential privacy generative adversarial networks machine learning |
| url | https://www.mdpi.com/2079-9292/9/5/874 |
| work_keys_str_mv | AT taehoonkim selectivefeatureanonymizationforprivacypreservingimagedatapublishing AT jihoonyang selectivefeatureanonymizationforprivacypreservingimagedatapublishing |