Deep Serial Number: Computational Watermark for DNN Intellectual Property Protection
Document Type
Conference Proceeding
Publication Date
1-1-2023
Abstract
In this paper, we present DSN (Deep Serial Number), a simple yet effective watermarking algorithm designed specifically for deep neural networks (DNNs). Unlike traditional methods that incorporate identification signals into DNNs, our approach explores a novel Intellectual Property (IP) protection mechanism for DNNs, effectively thwarting adversaries from using stolen networks. Inspired by the success of serial numbers in safeguarding conventional software IP, we propose the first implementation of serial number embedding within DNNs. To achieve this, DSN is integrated into a knowledge distillation framework, in which a private teacher DNN is initially trained. Subsequently, its knowledge is distilled and imparted to a series of customized student DNNs. Each customer DNN functions correctly only upon input of a valid serial number. Experimental results across various applications demonstrate DSN’s efficacy in preventing unauthorized usage without compromising the original DNN performance. The experiments further show that DSN is resistant to different categories of watermark attacks.
Identifier
85174435483 (Scopus)
ISBN
[9783031434266]
Publication Title
Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics
External Full Text Location
https://doi.org/10.1007/978-3-031-43427-3_10
e-ISSN
16113349
ISSN
03029743
First Page
157
Last Page
173
Volume
14174 LNAI
Grant
CNS-1816497
Fund Ref
National Science Foundation
Recommended Citation
Tang, Ruixiang; Du, Mengnan; and Hu, Xia, "Deep Serial Number: Computational Watermark for DNN Intellectual Property Protection" (2023). Faculty Publications. 2100.
https://digitalcommons.njit.edu/fac_pubs/2100