IP Library Granted Patent US 12,499,187
Granted Patent B2
US 12,499,187 · App. 18/450,092 · Granted Dec 16, 2025

Methods and systems for watermarking digital data

Inventors: Saeed Ranjbar Alvar (Coquitlam, CA); Mohammad Akbari (Coquitlam, CA); Ming Xuan Yue (Richmond, CA); Yong Zhang (Richmond, CA)
Assignee: Huawei Cloud Computing Technologies Co., Ltd.
G06F21/16G06T1/0071H04L9/3247H04N21/23892H04N21/8358G06T2201/0062H04L2209/608
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,499,187
App. No.
18/450,092
Granted
Dec 16, 2025
Kind
B2
Abstract

Systems and methods for watermarking objects of a digital dataset are disclosed. The method includes, at a first moment in time, acquiring a watermark message of a first size and generating a plurality of sub-messages based on the watermark message. The plurality of sub-messages are indicative of encoded portions of the watermark message. Each of the plurality of sub-messages is different from the watermark message and having a second size, the second size being smaller than the first size. The method also includes generating a watermarked object by embedding a given object from the digital dataset with a given one of the plurality of sub-messages, in lieu of the watermark message.

Claims (70)

1 . A computer-implemented method for watermarking objects of a digital dataset, the method executable by one or more processors, the method comprising:

at a first moment in time:

acquiring a watermark message of a first size;

generating a plurality of sub-messages based on the watermark message, the plurality of sub-messages being indicative of encoded portions of the watermark message,

each of the plurality of sub-messages being different from the watermark message and having a second size, the second size being smaller than the first size;

wherein the generating the plurality of sub-messages includes:

generating a plurality of codewords using the watermark message; and

generating a first sub-message from the plurality of sub-messages using a combination of codewords from the plurality of codewords; and

generating a watermarked object by embedding a given object from the digital dataset with the first sub-message, in lieu of the watermark message.

2 . The method of claim 1 , wherein the method further comprises:

at a second moment in time after the first moment in time:

acquiring a test watermarked digital dataset including a given test watermarked object;

extracting a test sub-message from the given test watermarked object, the test sub-message having the second size;

generating a decoded watermark message using the test sub-message, the decoded watermark message being different from the test sub-message and having the first size; and

determining whether the decoded watermark message matches the watermark message.

3 . The method of claim 1 , wherein the given object is at least one of a text-based object, an audio object, and a video object.

4 . The method of claim 1 , wherein the watermark message includes at least one of textual data, audio data, and video data.

5 . The method of claim 1 , wherein the generating the plurality of sub-messages is executed by employing a pre-trained Neural Network (NN) model.

6 . The method of claim 1 , wherein

the first sub-message including:

a body portion with the combination of codewords, and

an indexing portion being unique for combination of codewords amongst other possible combinations of codewords from the plurality of codewords.

7 . The method of claim 6 , wherein the generating the plurality of sub-messages includes:

generating a second sub-message from the plurality of sub-messages using a second combination of codewords from the plurality of codewords, the second combination of codewords being different from the combination of codewords, the first sub-message including:

a second body portion with the second combination of codewords, and

a second indexing portion being unique for second combination of codewords amongst other possible combinations of codewords from the plurality of codewords;

generating a second watermarked object by embedding a second given object from the plurality of objects with the second sub-message, in lieu of the watermark message.

8 . A method of processing a watermarked digital dataset, the method executable by one or more processors, the method comprising:

at a first moment in time:

acquiring a watermark message of a first size;

generating a plurality of sub-messages based on the watermark message, the plurality of sub-messages being indicative of encoded portions of the watermark message,

each of the plurality of sub-messages being different from the watermark message and having a second size, the second size being smaller than the first size;

wherein the generating the plurality of sub-messages includes:

generating a plurality of codewords using the watermark message;

generating a first sub-message from the plurality of sub-messages using a combination of codewords from the plurality of codewords,

generating a watermarked object by embedding a given object from the digital dataset with the first sub-message, in lieu of the watermark message,

at a second moment in time after the first moment in time:

acquiring a test watermarked digital dataset including a given test watermarked object;

extracting a test sub-message from the given test watermarked object, the test sub-message having the second size;

generating a decoded watermark message using the test sub-message, the decoded watermark message being different from the test sub-message and having the first size, the first size being bigger than the second size; and

determining whether the decoded watermark message matches a stored watermark message.

9 . The method of claim 8 , wherein the method further comprises extracting a second test sub-message from a second test watermarked object from the test watermarked digital dataset and wherein the generating the decoded watermark message further includes using the second test sub-message.

10 . A system for watermarking objects of a digital dataset, the system comprising a controller and a memory storing a plurality of executable instructions which, when executed by the controller, cause the system to:

at a first moment in time:

acquire a watermark message of a first size;

generate a plurality of sub-messages based on the watermark message, the plurality of sub-messages being indicative of encoded portions of the watermark message,

each of the plurality of sub-messages being different from the watermark message and having a second size,

the second size being smaller than the first size;

wherein the system is configured to generate the plurality of sub-messges by:

generating a plurality of codewords using the watermark message; and

generating a first sub-message from the plurality of sub-messages using a combination of codewords from the plurality of codewords;

generate a watermarked object by embedding a given object from the digital dataset with the first sub-message, in lieu of the watermark message.

11 . The system of claim 10 , further configured to:

at a second moment in time after the first moment in time:

acquire a test watermarked digital dataset including a given test watermarked object;

extract a test sub-message from the given test watermarked object, the test sub-message having the second size;

generate a decoded watermark message using the test sub-message, the decoded watermark message being different from the test sub-message and having the first size; and

determining whether the decoded watermark message matches the watermark message.

12 . The system of claim 10 , wherein the given object is at least one of a text-based object, an audio object, and a video object.

13 . The system of claim 10 , wherein the watermark message includes at least one of textual data, audio data, and video data.

14 . The system of claim 10 , wherein the generating the plurality of sub-messages is executed by employing a pre-trained Neural Network (NN) model.

15 . The system of claim 10 , wherein

the first sub-message including:

a body portion with the combination of codewords, and

an indexing portion being unique for combination of codewords amongst other possible combinations of codewords from the plurality of codewords.

16 . The system of claim 15 , wherein the system is configured to generate the plurality of sub-messages by:

generating a second sub-message from the plurality of sub-messages using a second combination of codewords from the plurality of codewords, the second combination of codewords being different from the combination of codewords, the first sub-message including:

a second body portion with the second combination of codewords, and

a second indexing portion being unique for second combination of codewords amongst other possible combinations of codewords from the plurality of codewords;

generating a second watermarked object by embedding a second given object from the plurality of objects with the second sub-message, in lieu of the watermark message.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 15, 2023
From: RANJBAR ALVAR, SAEED; AKBARI, MOHAMMAD; YUE, MING XUAN; ZHANG, YONG
To: HUAWEI CLOUD COMPUTING TECHNOLOGIES CO., LTD.
Reel/Frame 064922/0873 →
Continuity (1)
Related Publication 20250062911A1 · Feb 20, 2025
References Cited (16)
US 6449380B1 · Acharya · 2002 [cited by examiner]
US 9911171B2 · Malik et al. · 2018 [cited by applicant]
US 11669601B2 · Jalali · 2023 [cited by examiner]
US 20030118208A1 · Epstein · 2003 [cited by applicant]
US 20030204727A1 · Sasaki · 2003 [cited by applicant]
US 20220130008A1 · Jalali · 2022 [cited by examiner]
US 20240087075A1 · Luo · 2024 [cited by examiner]
Sablayrolles et al., “Radioactive data: tracing through training”, In International Conference on Machine Learning, Nov. 21, 2020, PMLR. [cited by applicant]
Wenger et al., “Data Isotopes for Data Provenance in DNNs”, arXiv preprint arXiv:2208.13893, Aug. 29, 2022. [cited by applicant]
Li et al., “Open-sourced dataset protection via backdoor watermarking”, NeurIPS 2020 Workshop on Dataset Curation and Security, Nov. 19, 2020. [cited by applicant]
Lu et al., “A hierarchical protection scheme for intellectual property of semi-open source datasets based on double watermarking”, Optik—Intl. J. for Light and Electron Opt., (2022), 269:169931. [cited by applicant]
Shehab et al., “Watermarking Relational Databases Using Optimization-Based Techniques”, IEEE transactions on Knowledge and Data Engineering. Jan. 2008, 20(1), pp. 116-129. [cited by applicant]
Thanh et al., “Pseudo Zero-watermarking Technique based on non-blind watermarking and VSS.”, Multimedia Tools and Applications, Apr. 2, 2022. [cited by applicant]
Tekgul et al., “On the Effectiveness of Dataset Watermarking in Adversal Settings”, in Proceedings of the 2022 ACM on International Workshop on Security and Privacy Analytics, Feb. 25, 2022. [cited by applicant]
Fernandez et al., “Watermarking Images in Self-Supervised Latent Spaces”, Mar. 23, 2022. [cited by applicant]
Li et al., “Untargeted Backdoor Watermark: Towards Harmless and Stealthy Dataset Copyright Protection”, 36th Conference on Neural Information Processing Systems (NeurIPS 2022), Apr. 5, 2023. [cited by applicant]