June 11th 2021

Research project on practicality of machine learning on encrypted data

Austrian consortium aims to enable cross-company usage of sensitive data


Research project on practicality of machine learning on encrypted data. Austrian consortium aims to enable cross-company usage of sensitive data.
Dr. Pascal Schöttle, Associate Professor and Martin Nocker, MSc, Research Assistant at MCI are part of the Austrian research team investigating the practicality of machine learning on encrypted data. Photo credit: Arno Rottensteiner


The project "Secure Machine Learning Applications with Homomorphically Encrypted Data" (SMiLe) is funded under project number 886524 by the Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology (BMK) as part of the 8th call of the "ICT of the Future" program.

Project partner:

MCI | The Entrepreneurial School: Assoc. Prof. Dr. Pascal Schöttle, Associate Professor; Martin Nocker, MSc, Research Assistant

Fraunhofer Austria

Software Competence Center Hagenberg

Center for Virtual Reality and Visualization

Companies: Fill, CORE smartwork and Tributech

Project duration:

The project has been running since April 2021 and will end in September 2023 after a duration of 30 months.

Status quo:

There is enormous potential in the data that is being collected and stored by companies in ever-increasing quantities, which can be tapped using machine learning methods. Since, on the one hand, machine learning works particularly well with large volumes of data, but, on the other hand, it is not worthwhile for every company to build up know-how and infrastructure for the application of corresponding processes, collaboration is an option for companies. When it comes to sensitive data, however, this is only an option for many companies if appropriate security precautions are taken. Pascal Schöttle, Associate Professor at MCI, believes that the importance of security in connection with machine learning will increase even further in the future. Research into methods that allow secure processing of data has been going on for years. Results to date show that a certain type of encryption, known as homomorphic encryption - is suitable, at least in theory - for enabling secure machine learning on sensitive data. In practice, however, know-how and suitable software are lacking. SMiLe addresses both aspects and thus aims to create an essential prerequisite for the practical use of machine learning on encrypted data.


The requirements with regard to the processing and protection of data are manifold. With Fill, CORE smartwork and Tributech, three companies are on board that have the most diverse requirements! The Austrian research team intends to investigate various potential applications during the course of the project. In order to successfully transfer the theoretical preliminary work into practice, a wide range of expertise is required: Fraunhofer Austria contributes experience with machine learning to the project. MCI provides the necessary knowledge about cryptographic methods. The Software Competence Center Hagenberg and the Center for Virtual Reality and Visualization also make important contributions to the project with their expertise in the areas of explainable artificial intelligence respectively data visualization.


In addition to insights into the practicality of homomorphic encryption, Pascal Schöttle expects the project to yield new theoretical findings. However, he is also aware of the challenges that need to be overcome: "One of the biggest challenges in the course of SMiLe will probably be to efficiently combine an exceedingly computationally intensive application, such as machine learning, with an equally computationally intensive form of encryption, as is the case with homomorphic encryption."

Pascal Schöttle is supported by Martin Nocker, Research Assistant at MCI, who also sees great potential in the SMiLe project: "Homomorphic encryption is referred to as the 'holy grail' of cryptography, as its properties allow sensitive data to be processed securely. We want to be part of the active and young community that continuously develops homomorphic encryption and combines the most suitable methods with state-of-the-art machine learning algorithms. In doing so, we must always pay attention to the best possible trade-off between security, accuracy and efficiency."