Scholarlink

With the increase in determining the relationshiop between genotypes and phenotypes, the use of third-party companies who provide services related to genotype-to-phenotype prediction has also increased. With this, the security of genotype data and models parameters must be preserved. In our work, we addressed this problem with homomorphic encryption. We developed a secure inference system using OpenFHE, leveraging threshold homomorphic encryption that enables the encryption of genotype data by clients and the encryption of model parameters by modelers using a shared public key. Logistic regression models were used for phenotype predictions. The evaluator could securely compute predictions without accessing plaintext data, and only clients could decrypt the final results.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Secure Phenotype Inference on Homomorphically Encrypted Genotype Data and Model Paremeters

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Dania Ashraf Mohamed,
  • Richard Bryann Chua

摘要

With the increase in determining the relationshiop between genotypes and phenotypes, the use of third-party companies who provide services related to genotype-to-phenotype prediction has also increased. With this, the security of genotype data and models parameters must be preserved. In our work, we addressed this problem with homomorphic encryption. We developed a secure inference system using OpenFHE, leveraging threshold homomorphic encryption that enables the encryption of genotype data by clients and the encryption of model parameters by modelers using a shared public key. Logistic regression models were used for phenotype predictions. The evaluator could securely compute predictions without accessing plaintext data, and only clients could decrypt the final results.