東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back to Search results for [ null ]

Switch To: Labeled | MARC Mode | ISBD

Preserving Data and Model Privacy in...

Jia, Qi.

Linked to FindBook

Google Book

Amazon

博客來

Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications./
Author:	Jia, Qi.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2019,
Description:	107 p.
Notes:	Source: Dissertations Abstracts International, Volume: 81-04, Section: B.
Contained By:	Dissertations Abstracts International81-04B.
Subject:	Computer engineering. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13898175
ISBN:	9781088304884

Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications.
Jia, Qi.

Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications. - Ann Arbor : ProQuest Dissertations & Theses, 2019 - 107 p.

Source: Dissertations Abstracts International, Volume: 81-04, Section: B.

Thesis (Ph.D.)--State University of New York at Binghamton, 2019.

This item must not be sold to any third party vendors.

With the dramatic growth of data in both amount and scale, distributed machine learning has become an important tool for the massive data to finish the tasks as prediction, classification, recognition, etc. However, due to the practical physical constraints and the potential privacy leakage of data, it is irrational to aggregate raw data from all data owners for learning or classifying purpose in an open distributed system. Therefore, the distributed privacy-preserving approaches become to significant roles to tackle such problems.Different to traditional secure tasks that only focus on protecting the data communication, the privacy preservation of distributed machine learning have three major challenges: data privacy, model privacy, and system design. Data privacy requires raw data should not be exposed to any other entities than data owner during the whole learning or classifying process. The violating of data privacy may cause the sensitive information lost for individual distributed users, such as medical records, personal profile, etc. Model privacy means that the well-learned machine learning model from learning party should not be revealed to any other parties or users. The violating of model privacy may result to intellectual property lost or system breaches in training party, for example a direct face recognition model expose can result to the lost of training face images. The system design means the special distributed structure should be considered in developing of privacy-preserving system. Inefficient designs of privacy-preserving algorithms can bring extensive computational burden to the system due to complex nodes connections. This becomes even severer if the cryptographic tools are applied.Regarding to these challenges, we conduct analysis, studies, and experiments to propose new privacy-preserving approaches. Regarding to the data privacy, we propose a new privacy-preserving data classification and similarity evaluation approach for distributed systems, where both the test data and trained model privacy is preserved during the data classifications in a distributed system. For model privacy, we propose our countermeasures to the targeted misclassification attack in black-box adversarial settings, where the model is preserved from the misclassification attacks. For system architecture, we propose an efficient privacy-preserving machine learning approach in hierarchical distributed system, where the normal learning methods are improved to efficiently run in the complicated hierarchical system architecture. Our researches cover these topics from the data distribution, system architecture, attack method, to learning process, classification scheme, and privacy preservation. Theorem and experiment analysis are provided to demonstrate the correctness, efficiency, effectiveness of our work.

ISBN: 9781088304884Subjects--Topical Terms:

621879
Computer engineering.

Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications.
LDR:03943nmm a2200325 4500 001 2264563
005 20200504070446.5
008 220629s2019 ||||||||||||||||| ||eng d
020 $a 9781088304884
035 $a (MiAaPQ)AAI13898175
035 $a AAI13898175
040 $a MiAaPQ $c MiAaPQ
100 1 $a Jia, Qi. $3 3541688
245 1 0 $a Preserving Data and Model Privacy in Distributed System for Machine Learning Based Classifications.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2019
300 $a 107 p.
500 $a Source: Dissertations Abstracts International, Volume: 81-04, Section: B.
500 $a Includes supplementary digital materials.
500 $a Advisor: Guo, Linke.
502 $a Thesis (Ph.D.)--State University of New York at Binghamton, 2019.
506 $a This item must not be sold to any third party vendors.
520 $a With the dramatic growth of data in both amount and scale, distributed machine learning has become an important tool for the massive data to finish the tasks as prediction, classification, recognition, etc. However, due to the practical physical constraints and the potential privacy leakage of data, it is irrational to aggregate raw data from all data owners for learning or classifying purpose in an open distributed system. Therefore, the distributed privacy-preserving approaches become to significant roles to tackle such problems.Different to traditional secure tasks that only focus on protecting the data communication, the privacy preservation of distributed machine learning have three major challenges: data privacy, model privacy, and system design. Data privacy requires raw data should not be exposed to any other entities than data owner during the whole learning or classifying process. The violating of data privacy may cause the sensitive information lost for individual distributed users, such as medical records, personal profile, etc. Model privacy means that the well-learned machine learning model from learning party should not be revealed to any other parties or users. The violating of model privacy may result to intellectual property lost or system breaches in training party, for example a direct face recognition model expose can result to the lost of training face images. The system design means the special distributed structure should be considered in developing of privacy-preserving system. Inefficient designs of privacy-preserving algorithms can bring extensive computational burden to the system due to complex nodes connections. This becomes even severer if the cryptographic tools are applied.Regarding to these challenges, we conduct analysis, studies, and experiments to propose new privacy-preserving approaches. Regarding to the data privacy, we propose a new privacy-preserving data classification and similarity evaluation approach for distributed systems, where both the test data and trained model privacy is preserved during the data classifications in a distributed system. For model privacy, we propose our countermeasures to the targeted misclassification attack in black-box adversarial settings, where the model is preserved from the misclassification attacks. For system architecture, we propose an efficient privacy-preserving machine learning approach in hierarchical distributed system, where the normal learning methods are improved to efficiently run in the complicated hierarchical system architecture. Our researches cover these topics from the data distribution, system architecture, attack method, to learning process, classification scheme, and privacy preservation. Theorem and experiment analysis are provided to demonstrate the correctness, efficiency, effectiveness of our work.
590 $a School code: 0792.
650 4 $a Computer engineering. $3 621879
650 4 $a Computer science. $3 523869
650 4 $a Electrical engineering. $3 649834
650 4 $a Data analysis. $2 bisacsh $3 3515250
690 $a 0464
690 $a 0984
690 $a 0544
710 2 $a State University of New York at Binghamton. $b Electrical and Computer Engineering. $3 3174435
773 0 $t Dissertations Abstracts International $g 81-04B.
790 $a 0792
791 $a Ph.D.
792 $a 2019
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13898175