| Περιγραφή: |
DNA data contains sensitive health information and personally identifiable data. Currently, even if DNA data is stored in encrypted databases, it must be decrypted for health professionals and researchers to analyze, which means that DNA data exists in plaintext on unsecured, untrusted servers and machines during analysis. This thesis describes a complete system for homomorphically encrypting DNA data in a trusted context and then running analytic operations on the encrypted DNA data in an untrusted context, thus allowing healthcare professionals and researchers to run both high volume analytics on many individuals’ sequenced DNA and run complex analytics on a single individual’s sequenced DNA without ever handling plaintext data. Symmetric encryption is used as a mechanism for controlling which queries are made on the data. The threat model addressed by this system allows an authorized party to run only authorized queries on a genome, while restricting any additional access. The system implemented achieves substring search, substring search with wildcards representing mutations, and percent match between two nucleotide sequences by converting genomic data into one-hot binary matrixes and encrypting each bit individually using OpenFHE’s LWE Encryption implemented using the CGGI scheme. While runtime for each operation is O(nm), each operation is maximally parallelized using OpenMP, thus allowing for accelerated performance on machines with multiple CPUs without the need for batching. |