iDASH secure genome analysis competition 2018: blockchain genomic data access logging, homomorphic encryption on GWAS, and DNA segment searching

Tsung-Ting Kuo, Xiaoqian Jiang, Haixu Tang, Xiaofeng Wang, Tyler Bath, Diyue Bu, Lei Wang, Arif Harmanci, Shaojie Zhang, Degui Zhi, Heidi J. Sofia, Lucila Ohno-MachadoBMC 2020 · DOI · OpenAlex

Abstract

This paper summarizes the results of the 2018 iDASH Secure Genome Analysis Competition, a premier event that challenges the global research community to solve the "privacy vs. utility" paradox in genomics. As genomic data becomes more accessible, the risk of re-identification and data misuse grows. This competition focused on three cutting-edge technologies—blockchain, homomorphic encryption, and secure multi-party computation—to prove that we can analyze sensitive DNA data without ever actually "seeing" it.

Introduction

Genome privacy presents a unique challenge: unlike a credit card number, DNA cannot be changed, and a breach affects not just the individual but their biological relatives. However, overly restrictive data silos stall medical progress. The 2018 iDASH competition addressed three critical gaps in current genomic data management:

  1. Auditability: Current repositories rely on local logging. If a central server is compromised, logs can be altered to hide unauthorized access.
  2. Scalability of Encrypted Computation: Homomorphic Encryption (HE) allows computation on encrypted data but has historically been too slow for large-scale Genome-Wide Association Studies (GWAS).
  3. Secure Relative Finding: Searching for relatives in large databases currently requires sharing raw genetic data with third parties, creating significant privacy risks.

Methods and Competition Tracks

The competition was divided into three distinct tracks, each targeting a specific privacy-preserving technology.

Track 1: Blockchain for Auditing

Teams utilized MultiChain (a Bitcoin fork) to create a decentralized, immutable log of data access. The objective was to ensure that every instance of a researcher accessing a dataset is permanently recorded across multiple nodes, preventing "log scrubbing" by malicious actors.

The proposed blockchain-based decentralized logging architecture for genomic data access auditing across multiple cloud environments.

Track 2: Homomorphic Encryption for GWAS

Participants used the HEAAN/CKKS encryption scheme. Unlike older methods requiring exact bit-wise operations, CKKS allows for "approximate arithmetic," which is significantly faster for the floating-point mathematics required in statistical analysis. Teams implemented semi-parallel algorithms to compute p-values for thousands of SNPs simultaneously.

Track 3: Secure DNA Segment Searching

This track focused on Secure Two-Party Computation (2PC). A "querier" and a "database owner" collaborated to find long, matching DNA segments (set-maximal matches) without either party revealing their underlying sequences to the other.

Key Insights and Results

The "CKKS Consensus"

A major takeaway was the convergence of winning teams on the CKKS/HEAAN framework. This suggests a community consensus that approximate arithmetic is the optimal path for privacy-preserving machine learning and statistics in genomics.

Performance Benchmarks

  • GWAS Speed: In Track 2, the winning teams (Duality Inc and UCSD) achieved high-accuracy results for 15,000 SNPs in under two minutes. Previously, processing a single SNP could take minutes.
  • Blockchain Scalability: The top solution (GersteinLab) successfully stored 800,000 genomic access records within one hour and supported complex Boolean queries in less than three minutes.
  • The Accuracy-Speed Tradeoff: In Track 3, results were mixed. While some solutions were fast (7 seconds), they lacked accuracy. The most accurate solution (CNRS) required 26 hours, indicating that secure relative-finding at scale remains a significant hurdle.
Overall competition scores across different participating teams and data scales (Small, Medium, Large).

Discussion and Future Work

The 2018 iDASH competition demonstrated that privacy-preserving genomic analysis is moving from theory to practice. However, limitations remain. While 15,000 SNPs is a breakthrough, real-world GWAS often involve millions of markers. Scaling HE to that level without massive hardware requirements is the next frontier.

Additionally, the blockchain track highlighted that while "on-chain" storage provides superior security, it currently incurs a significant storage penalty (up to 7.8 GB for 800k records). Future research will likely focus on hybrid approaches—combining secure hardware like Intel SGX with software-based encryption—to bridge the performance gap for real-time consumer applications.

Summary updated

BibTeX
@inproceedings{kuo2020idash,
  author = {Tsung-Ting Kuo and Xiaoqian Jiang and Haixu Tang and Xiaofeng Wang and Tyler Bath and Diyue Bu and Lei Wang and Arif Harmanci and Shaojie Zhang and Degui Zhi and Heidi J. Sofia and Lucila Ohno-Machado},
  title = {{iDASH secure genome analysis competition 2018: blockchain genomic data access logging, homomorphic encryption on GWAS, and DNA segment searching}},
  booktitle = {BMC},
  year = {2020},
  doi = {10.1186/s12920-020-0715-0},
}
Figure