Privacy-Preserving Fingerprinting Against Collusion and Correlation Threats in Genomic Data.

Area of Science:

• Computational biology
• Genomic data security
• Bioinformatics

Background:

• Sharing genomic databases is crucial for collaborative research, but raises concerns about unauthorized redistribution and participant privacy.
• Existing methods struggle to protect sensitive genomic data against collusion and correlation attacks.
• Genome-wide association studies (GWAS) rely on accurate, shared data, necessitating robust protection mechanisms.

Purpose of the Study:

• To develop a novel genomic database fingerprinting scheme, Gen-Scope, that ensures copyright protection and privacy preservation.
• To assess the robustness of Gen-Scope against collusion and correlation attacks.
• To improve the accuracy and utility of shared genomic databases for research.

Main Methods:

• Developed Gen-Scope, a novel fingerprinting scheme combining copyright protection (traceability) and privacy preservation (local differential privacy).
• Augmented Gen-Scope with Tardos codes for robust traitor tracing against collusion attacks.
• Evaluated Gen-Scope using a real-world genomic database, testing its resilience to data alteration and collusive behavior.

Main Results:

• Gen-Scope demonstrates strong fingerprint robustness, withstanding up to 45% data alteration by attackers.
• The scheme effectively detects colluders with high probability, even after independent correlation attacks.
• Gen-Scope outperforms baseline methods, achieving ~10% higher accuracy and ~20% greater consistency in variant-phenotype associations.

Conclusions:

• Gen-Scope offers a robust solution for secure genomic database sharing, balancing copyright protection with privacy preservation.
• The developed fingerprinting scheme significantly enhances the security and reliability of shared genomic data for computational biology research.
• Gen-Scope's ability to identify colluders under combined attacks marks a significant advancement in protecting sensitive biological datasets.