Dataset Fingerprinting via Cluster-Based Watermark Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data watermarking and fingerprinting techniques are easily detectable and can be reverse-engineered, posing risks to data privacy and traceability, especially in shared datasets.
Innovation Solution
A method involving clustering datasets using a secret key and security parameters to embed fingerprint data randomly across records, making it difficult to detect without the secret key, ensuring improved traceability and ownership verification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional fingerprinting and watermarking techniques are applied to datasets, then data traceability and ownership verification are improved, but the watermarking becomes easily detectable and can be reverse-engineered by malicious actors
Solution Approach 1:
The patent changes the parameters of watermarking by using clustering-based selection instead of fixed-position embedding. Records are selected for watermarking based on their cluster assignments and security parameters, making the watermarking pattern dynamic and non-reproducible without the secret key, thus preventing easy detection and reverse-engineering while maintaining traceability
Solution Approach 2:
The patent introduces clustering as an intermediary mechanism between the watermarking system and the dataset. The clustering process, guided by security parameters and secret keys, acts as a mediator that selects which records receive watermarks, creating a complex dependency chain that prevents direct analysis and reverse-engineering of the watermarking pattern
2Object-affected harmful factors
If a few records in each portion are replaced with the same watermarking data, then data privacy protection is achieved, but the watermarking can be easily extracted and used to generate clean datasets
Solution Approach 1:
The patent introduces asymmetry in the watermarking process by using different selection criteria for different records based on their cluster assignments. Not all records are treated equally; instead, records are selectively watermarked based on security parameters and cluster characteristics, creating an asymmetric pattern that is difficult to predict and extract without the secret key
Solution Approach 2:
The patent makes the watermarking process dynamic by using clustering that can adapt to different dataset characteristics and security parameters. The selection of records for watermarking is not fixed but depends on the clustering results and security parameters, creating a dynamic watermarking pattern that is hard to reverse-engineer
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Aspects of the present disclosure provide systems, methods, apparatus, and computer-readable storage media that support improved watermarking and fingerprinting of a shared dataset. To illustrate, clustering may be performed on the dataset using initial clustering parameters (e.g., a secret key) to assign each record (e.g., attribute) of the dataset to one of multiple clusters. The secret key may be selected by a user or determined automatically based on the clustering algorithm. After the clustering, the records of each cluster may be selected for embedding a portion of fingerprint data based on one or more security parameters (e.g., a hash function, priority values, even/or selection, etc.). The selected records (or portions thereof) may be replaced with corresponding portions of the fingerprint data to embed the fingerprint data within different records as watermarking. Aspects also include analyzing a dataset to verify whether watermarking is present and to extract a fingerprint.