Andreas Trügler

Details about my work and research.


Privacy-preserving analytics and computation on encrypted data

Andreas Trügler

2-Minute Read


Encryption and privacy

Due to my current position as co-head of the data privacy for AI research area at Know-Center and my focus on privacy-enhancing technologies (see projects), I’m very happy about the opportunity to collaborate with Christian Rechberger and his excellent team, renowned experts in the field of cryptography and privacy.

The idea of identifying a problem that is mathematically hard to solve and using it as basis for secure communication has always been intriguing to me. Ever since I counted the letter frequencies in Poe’s Gold-Bug to find the pirate treasure of the infamous Captain Kidd I was hooked to the topic and I’m glad my chosen career path allows me to contribute a little bit to this interesting field of research. Leaving pirate treasures behind modern cryptography has of course little to do with such romanticism, it is the mathematical foundation of everything secure and also a central pillar of our financial, economic and societal system. The right to privacy is anchored in the Universal Declaration of Human Rights and especially in times like these where large companies more and more emerge as data monopolists it is of tremendous importance to think about privacy.

Homomorphic encryption and machine learning

In order to use the full capacity of machine learning and artificial intelligence we need corresponding data sets for training and evaluation. Usually the more data the merrier regarding our predictions and simulations. In newspaper articles or during public discussions you often get the impression that you have to choose here, either you want to have the full functionality of machine learning and you then have to give up some privacy for that or you keep your data to yourself, pull the plug and resign from the capabilities of data-driven evaluations. Such a wording is simply wrong, modern cryptographic methods like mutliparty computation or homomorphic encryption allow to perfrom calculations on encrypted data.

With homomorphic encryption encrypted data can be evaluated on a server without revealing the true content.

With homomorphic encryption encrypted data can be evaluated on a server without revealing the true content.

Thus without giving up any privacy you could still combine medical records of different hospitals to train a mutual disease predicting algorithm, companies could share their models with suppliers or collaborateurs without revealing any company secrets or homomorphic encryption could also be used to generate a privacy-preserving heatmap of Covid infections.

Current projects: DDAI and TRUSTS

The main focus of my work at Know-Center is currently centered around the DDAI Comet module and the Horizon 2020 project TRUSTS. Both projects focus on privacy-preserving analytics based on maschine learning and cryptography.

Research Interests


I'm a scientist and researcher working on privacy-preserving machine learning applications.