Differential privacy and membership inference attacks

The growing use of machine learning on various datasets results in privacy concerns about records of the data being leaked. Membership inference is a type of attack that identifies the members of the training dataset. The research studies a privacy-preserving mechanism, differential privacy, to miti...

Full description

Saved in:
Bibliographic Details
Main Author: Ong, Ting Yu
Other Authors: Wang Huaxiong
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2023
Subjects:
Online Access:https://hdl.handle.net/10356/166457
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-166457
record_format dspace
spelling sg-ntu-dr.10356-1664572023-05-08T15:39:17Z Differential privacy and membership inference attacks Ong, Ting Yu Wang Huaxiong School of Physical and Mathematical Sciences Institute for Infocomm Research Benjamin Tan Hong Meng HXWang@ntu.edu.sg, benjamin_tan@i2r.a-star.edu.sg Science::Mathematics::Applied mathematics The growing use of machine learning on various datasets results in privacy concerns about records of the data being leaked. Membership inference is a type of attack that identifies the members of the training dataset. The research studies a privacy-preserving mechanism, differential privacy, to mitigate membership inference attacks. Generally, there is a lack of studies that include the two mentioned concepts: membership inference and differential privacy. This research extends the concepts to the less-tested datasets to understand the interaction between the concepts. Image, Time Series and Natural Language Processing datasets were used to train the target models and the reference models. As expected, differential privacy does hinder the membership inference attack by reducing it to a random guess for Image Dataset. However, for the other types of data, there are no observable changes before and after the implementation of differential privacy. Hence, the implementation of differential privacy was able to maintain the attack at a random guess level, suggesting that implementing differential privacy can help to mitigate the membership inference attack. Bachelor of Science in Mathematical Sciences 2023-05-02T02:52:40Z 2023-05-02T02:52:40Z 2023 Final Year Project (FYP) Ong, T. Y. (2023). Differential privacy and membership inference attacks. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/166457 https://hdl.handle.net/10356/166457 en application/pdf Nanyang Technological University
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Science::Mathematics::Applied mathematics
spellingShingle Science::Mathematics::Applied mathematics
Ong, Ting Yu
Differential privacy and membership inference attacks
description The growing use of machine learning on various datasets results in privacy concerns about records of the data being leaked. Membership inference is a type of attack that identifies the members of the training dataset. The research studies a privacy-preserving mechanism, differential privacy, to mitigate membership inference attacks. Generally, there is a lack of studies that include the two mentioned concepts: membership inference and differential privacy. This research extends the concepts to the less-tested datasets to understand the interaction between the concepts. Image, Time Series and Natural Language Processing datasets were used to train the target models and the reference models. As expected, differential privacy does hinder the membership inference attack by reducing it to a random guess for Image Dataset. However, for the other types of data, there are no observable changes before and after the implementation of differential privacy. Hence, the implementation of differential privacy was able to maintain the attack at a random guess level, suggesting that implementing differential privacy can help to mitigate the membership inference attack.
author2 Wang Huaxiong
author_facet Wang Huaxiong
Ong, Ting Yu
format Final Year Project
author Ong, Ting Yu
author_sort Ong, Ting Yu
title Differential privacy and membership inference attacks
title_short Differential privacy and membership inference attacks
title_full Differential privacy and membership inference attacks
title_fullStr Differential privacy and membership inference attacks
title_full_unstemmed Differential privacy and membership inference attacks
title_sort differential privacy and membership inference attacks
publisher Nanyang Technological University
publishDate 2023
url https://hdl.handle.net/10356/166457
_version_ 1770567281154195456