Please use this identifier to cite or link to this item:
http://hdl.handle.net/10397/88239
DC Field | Value | Language |
---|---|---|
dc.contributor | Department of Electronic and Information Engineering | en_US |
dc.creator | Zheng, H | en_US |
dc.creator | Ye, Q | en_US |
dc.creator | Hu, H | en_US |
dc.creator | Fang, C | en_US |
dc.creator | Shi, J | en_US |
dc.date.accessioned | 2020-09-28T08:33:27Z | - |
dc.date.available | 2020-09-28T08:33:27Z | - |
dc.identifier.isbn | 978-3-030-29958-3 (print) | en_US |
dc.identifier.isbn | 978-3-030-29959-0 (online) | en_US |
dc.identifier.issn | 0302-9743 | en_US |
dc.identifier.uri | http://hdl.handle.net/10397/88239 | - |
dc.description | 24th European Symposium on Research in Computer Security, Luxembourg, September 23-27, 2019 | en_US |
dc.language.iso | en | en_US |
dc.publisher | Springer | en_US |
dc.rights | © Springer Nature Switzerland AG 2019 | en_US |
dc.rights | Zheng H., Ye Q., Hu H., Fang C., Shi J. (2019) BDPL: A Boundary Differentially Private Layer Against Machine Learning Model Extraction Attacks. In: Sako K., Schneider S., Ryan P. (eds) Computer Security – ESORICS 2019. ESORICS 2019. Lecture Notes in Computer Science, vol 11735. Springer, Cham. | en_US |
dc.rights | The final authenticated version is available online at https://doi.org/10.1007/978-3-030-29959-0_4. | en_US |
dc.title | BDPL : a boundary differentially private layer against machine learning model extraction attacks | en_US |
dc.type | Conference Paper | en_US |
dc.identifier.spage | 66 | en_US |
dc.identifier.epage | 83 | en_US |
dc.identifier.volume | 11735 | en_US |
dc.identifier.doi | 10.1007/978-3-030-29959-0_4 | en_US |
dcterms.abstract | Machine learning models trained by large volume of proprietary data and intensive computational resources are valuable assets of their owners, who merchandise these models to third-party users through prediction service API. However, existing literature shows that model parameters are vulnerable to extraction attacks which accumulate a large number of prediction queries and their responses to train a replica model. As countermeasures, researchers have proposed to reduce the rich API output, such as hiding the precise confidence level of the prediction response. Nonetheless, even with response being only one bit, an adversary can still exploit fine-tuned queries with differential property to infer the decision boundary of the underlying model. In this paper, we propose boundary differential privacy ( ϵ -BDP) as a solution to protect against such attacks by obfuscating the prediction responses near the decision boundary. ϵ -BDP guarantees an adversary cannot learn the decision boundary by a predefined precision no matter how many queries are issued to the prediction API. We design and prove a perturbation algorithm called boundary randomized response that can achieve ϵ -BDP. The effectiveness and high utility of our solution against model extraction attacks are verified by extensive experiments on both linear and non-linear models. | en_US |
dcterms.accessRights | open access | en_US |
dcterms.bibliographicCitation | Lecture notes in computer science (including subseries Lecture notes in artificial intelligence and lecture notes in bioinformatics), 2019, v. 11735, p. 66-83 | en_US |
dcterms.isPartOf | Lecture notes in computer science (including subseries Lecture notes in artificial intelligence and lecture notes in bioinformatics) | en_US |
dcterms.issued | 2019 | - |
dc.relation.conference | European Symposium on Research in Computer Security [ESORICS] | en_US |
dc.identifier.eissn | 1611-3349 | en_US |
dc.description.validate | bcrc 202009 | en_US |
dc.description.oa | Accepted Manuscript | en_US |
dc.identifier.FolderNumber | a0483-n01 | - |
dc.description.pubStatus | Published | en_US |
Appears in Collections: | Conference Paper |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
ESORICS19.pdf | Pre-Published version | 1.23 MB | Adobe PDF | View/Open |
Page views
365
Last Week
3
3
Last month
Citations as of Mar 24, 2024
Downloads
495
Citations as of Mar 24, 2024
SCOPUSTM
Citations
30
Citations as of Mar 29, 2024
WEB OF SCIENCETM
Citations
19
Citations as of Mar 28, 2024
Google ScholarTM
Check
Altmetric
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.