Google has published an open set of libraries and utilities with the implementation of a full homomorphic encryption system that allows you to process data in encrypted form that do not appear in open form at any stage of the calculation. The toolkit makes it possible to create programs for confidential computing that can work with data without decryption, including performing mathematical and simple string operations on encrypted data. The project code is written in C++ and distributed under the Apache 2.0 license.
Unlike end-to-end encryption, homomorphic encryption, in addition to protecting data transmission, provides the ability to process data without decrypting it. Full homomorphism refers to the ability to perform addition and multiplication of encrypted data, starting from which any arbitrary calculations can be implemented. The output is an encrypted result, which would be similar to the encryption of the result of similar operations on the original data.
Working with data with homomorphic encryption comes down to the fact that the user encrypts the data and transfers it to a third-party service for processing without revealing the keys. This service performs the declared calculations and generates an encrypted result, without being able to determine what kind of data it works with. The user, using his keys, decrypts the given data and receives the result in clear text.
From the areas of application of homomorphic encryption, the creation of cloud services for confidential computing, the implementation of electronic voting systems, the creation of anonymized routing protocols, the processing of requests for encrypted data in a DBMS, and the confidential training of machine learning systems are noted.
For example, homomorphic encryption will be useful in medical applications that can receive sensitive information from patients in encrypted form and allow medical professionals to conduct analytics and detect deviations without decryption. Homomorphic encryption can also help in research that studies the relationship between diseases and certain genetic mutations, which requires the analysis of thousands of samples of genetic information.
A distinctive feature of the published toolkit is the ability to create programs for processing encrypted data using typical C++ development techniques. Using the provided transpiler, a C++ program is converted into a special FHE-C++ dialect capable of working with encrypted data.
Source: opennet.ru