Updated on 2026-09-10 GMT+08:00

Data Watermarking Overview

If data leakage occurs, you can use DSC to extract the watermark information. In this case, the organization or person that is accountable for the leakage problem can be easily found. Adding watermarks does not affect the distributed data usage.

Scenarios

Data watermarking is widely used in government departments, healthcare agencies, financial institutions, academic institutes, and other organizations. It is generally used for copyright protection and source tracing.

Data copyright protection: When digital assets are downloaded, copied, or provisioned to third parties for data mining, embedded watermarks can help to resolve copyright disputes.
Source tracing: Data provided for internal employees or third parties can be injected with watermarks to identify the ownership and remind them of keeping the data secure. When the data leaked, the watermarks can be used to trace the source of data leak and identify the root cause.

Advantages and Highlights

Visible and invisible watermarks: Embed visible or invisible watermarks into assets to effectively mitigate data theft via image manipulation, photos, or screenshots.
Detectable and tamper-proof: Embedded watermarks remain reliably detectable, providing robust resistance against loss, forgery, and tampering.
High robustness: Watermarks resist removal during transmission and active use. Even if the underlying data is tampered with or corrupted, the watermark remains highly extractable.

Product Capabilities

The following table describes data watermarking types and their capabilities.

**Table 1** Data watermark types supported by DSC
Watermark Type	Description	Data Sources Supporting Embedding/Extraction	Supported Format
Database watermark	Watermarks can be added to tables in a database. Lossless - pseudo column watermark: A pseudo column is generated based on attributes from the relational table, blending in seamlessly to evade attacker detection. Watermarks are embedded into this pseudo column to prevent distortion or modification of the original data. Lossless - pseudo row watermark: Pseudo rows are generated based on the existing data type, format, and value range. Watermarks are embedded into these pseudo rows to prevent distortion or modification of the original data. Lossy - column watermark: Watermarks are injected into existing column data, which modifies or alters the original values.	DWS	smallint, integer, bigint, float4, float8, varchar, text, and char
Database watermark		MRS-HIVE	smallint, int, long, float, double, and string
Document watermark	Visible or invisible data watermarks can be added to documents. Invisible watermarks do not affect the normal use of documents and are imperceptible to the human eye.	OBS buckets and local files	Documents: PDF, PPTX, DOCX, and XLSX
Document watermark		OBS buckets and local files	JSON data (data watermark APIs supported): integer, floating point, and string
Image watermark	Visible or invisible watermarks can be added to images. For example, you can overlay a transparent watermarked layer onto an image. This addition preserves the image's original visual utility and remains completely imperceptible to the human eye.	OBS buckets and local files	Images (image watermark API supported): JPG, JPEG, PNG, BMP, TIFF, TIF, TGA, and GIF

Architecture

Administrators preconfigure watermark policies in the data watermarking system, which supports both batch and real-time image watermarking. In batch processing mode, the system reads image files from OBS, embeds watermarks in batches, and saves the output back to OBS. In real-time processing mode, application or data-sharing systems pull raw images from OBS and invoke the watermark API to add visible or invisible watermarks in real time. Watermarked images can be distributed externally or shared with internal and third-party systems. If image data is leaked and transferred to unauthorized areas, the Internet, or external users, you can use API Gateway to connect to the data watermarking system and extract watermarks from leaked images. The watermark information can be used to trace the source and locate the data leakage entity. This enables source tracing and evidence collection for leakage incidents.

Click to enlarge

References

Injecting or Extracting watermarks from databases: Database watermarking embeds unique identifiers (including copyright claims, ownership information, and lineage tracking codes) directly into database records. This technology serves as a critical control for copyright protection, breach tracing, tamper detection, and compliance auditing, making it a foundational tool for data asset security and governance.
Injecting or extracting watermarks from documents: You can add or extract watermarks for local files or cloud files stored in OBS buckets.
Injecting or extracting watermarks from images: You can add custom watermarks to local images or cloud images stored in OBS buckets.
Watermark source tracing: DSC provides a unified entry for watermark source tracing.

Parent Topic: Data Watermarking