Privacy- and confidentiality-preserving AI/ML

In projects such as trups.eu and morphmetro.eu, we used Fully Homomorphic Encryption (FHE), which allows data to remain encrypted even during processing. This enables privacy- and confidentiality-preserving analytics and machine learning, without exposing any sensitive information.

For example, Laboratory A could send a patient's blood data to Laboratory B, which would perform the analysis without knowing the actual blood values or the outcome - because it is working with encrypted data. Only Laboratory A, as the data owner, can access both the input data (blood values) and the results of the analysis performed by Laboratory B, since only Laboratory A can decrypt the data.

Of course, this is just one example - many other use cases are possible (in life sciences, the financial industry, Industry 4.0, smart cities, smart grids, etc.), wherever data is analyzed by a third party - that is, in situations where it is necessary to ensure the privacy and confidentiality of the data being analyzed.