| Author: | Jiao, Guangyi |
| Title: | Model extraction and defenses on generative adversarial networks a comprehensive survey |
| Advisors: | Hu, Haibo (EEE) |
| Degree: | M.Sc. |
| Year: | 2024 |
| Department: | Department of Electrical and Electronic Engineering |
| Pages: | vi, 49 pages |
| Language: | English |
| Abstract: | Generative Adversarial Networks (GANs) have emerged as a powerful tool in machine learning, enabling tasks such as image synthesis, data augmentation, and style transfer. However, their capabilities and widespread deployment have also made them a target for model extraction attacks, where adversaries replicate the functionality or distribution of a target GAN through query-based interactions. This paper provides a comprehensive survey of GAN model extraction attacks and corresponding defense mechanisms. We classify extraction attacks into black-box, gray-box, and white-box scenarios, highlighting key techniques such as fidelity extraction, accuracy extraction, latent space interpolation, and domain shift mitigation. These methods demonstrate the feasibility of reconstructing GAN models even when operating with limited information or queries. In parallel, we examine existing defenses, including input/output perturbations, query limitations, and ownership verification. While these approaches offer some protection, they often degrade generation quality or can be circumvented by sophisticated attackers. The survey reveals an imbalance between the strength of extraction methods and the limitations of current defenses, emphasizing the need for more robust and adaptable security mechanisms. Finally, we outline future research directions, including improving defense techniques, expanding studies to domains such as text and audio, and exploring the economic feasibility of attacks and defenses. This work aims to provide a foundation for understanding the vulnerabilities of GANs and guide the development of secure and sustainable generative systems. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 8720.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 364.43 kB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14053