IND[i]GOME

The solution to securing genomic data is not more encryption. It is smarter encryption. It's IND[i]GOME. Data is digitally secured, no matter where you send or take it. Is your data in good hands?

Comment 1

Inspiration

Pure curiosity. The chance to attempt a difficult problem.

What it does

IND[i]GOME is designed to be a platform as a service (PaaS) available to companies and organizations doing genomic research and data collection. This tool implements fully-homomorphic encryption (FHE) of user data, allowing statistical analysis that can be built on top of a cryptographic scheme.

How we built it

We used an existing implementation, Fast Fully Homomorphic Encryption over the Torus (TFHE), geared toward developing applications. Our team applied fully homomorphic encryption (FHE) to sample data from the 1000 Genomes Project.

What is homomorphic encryption?

Homomorphic encryption is a form of encryption that allows computations to be performed on encrypted data without first having to decrypt it.

Challenges we ran into

Performance is terrible on the CPU. The literature depicts 20x-40x speedups on the GPU, which is the next step. This is one of the benefits of this solution: it can be easily parallelized to a GPU. A secure processor, which offers some level of the same privacy, is dependent on a limited power processor without a SIMT architecture. You're totally out of luck if you've got specialized FPGA solutions though.

Accomplishments that we're proud of

  • The synergy that was developed on our team in a short of time.
  • We feel that we inverted the problem to develop a creative solution.
  • Each IND[i]GOME team member was technically challenged to complete this product.
  • We produced a proof of concept that tackled a difficult and relevant problem.

What we learned

  • More about RUST as a programming language
  • Learned how to read and parse .vcf files
  • Learned how to use Streamlit
  • How to implement programs on top of homomorphic encryption

What's next for IND[i]GOME

  • Increasing speed in terms of parallelism and specialized hardware
  • Additional analysis functions on top of a homomorphic (dot product and convolutions)
  • Expanding into neural networks
  • Gaining more insights into the corner cases of genomics to build a more robust system

Built With

Share this project:

Updates