Privacy-preserving and robust watermarking on sequential genome data using belief propagation and local differential...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomic Data Security

Background:

• Genome sequencing is increasingly accessible, leading to widespread data sharing for medical and social purposes.
• Sharing genome data, even partially, poses significant privacy risks to donors.
• Unauthorized data sharing raises liability concerns, necessitating robust protection mechanisms.

Purpose of the Study:

• To develop a novel watermarking method for sequential genome data to detect unauthorized sharing.
• To ensure the robustness of watermarking against malicious adversaries and collusion attacks.
• To achieve epsilon-local differential privacy in genome data sharing with service providers.

Main Methods:

• A novel watermarking method utilizing the belief propagation algorithm for sequential genome data.
• Embedding robust watermarks to prevent tampering and ensure high detection probability.
• Incorporating publicly available genomic information (Minor Allele Frequency, Linkage Disequilibrium, Phenotype, Familial Information) for enhanced robustness.

Main Results:

• Achieved 100% detection rate against single service provider attacks with a 3% watermark length.
• Reached 80% detection with 5% watermark length and 90% detection with 10% watermark length in worst-case collusion attacks (50% malicious SPs).
• Demonstrated negligible impact of epsilon on precision, ensuring high privacy preservation.

Conclusions:

• The proposed watermarking scheme effectively detects unauthorized genome data sharing.
• The method provides robust protection against various attack scenarios, including collusion.
• High levels of privacy are maintained during data sharing with service providers.