Innovative technological methods allow sensitive information to be evaluated while keeping its content confidential. At first, it sounds like wishful thinking. However, the development of novel encryption protocols that are also safe from attacks by quantum computers are based on recent mathematical principles. Their application holds enormous potential.
How to keep data confidential and protected

In 1890, two American lawyers published an article entitled “The Right to Privacy”. It is considered one of the first major pleas for protection of personal data. In addition, this right has been preserved within the United Nations Universal Declaration of Human Rights for almost 75 years. Meanwhile, the technological prospects have changed to a great extent and due to digitization, the topic is more prevailing than ever. By means of the General Data Protection Regulation, a corresponding cornerstone was further expanded within the European Union. But how much privacy do you have to surrender to use digital services or the enormous possibilities of modern-day artificial intelligence (AI)? Are data protection and AI contradicting each other? Not at all since modern cryptographic methods allow using AI algorithms without the need of disclosing data.

Safe Algorithms

“Our goal is to create a new generation of artificial intelligence that follows ethical and trustworthy principles, that protects data content and delivers explainable and transparent results. Data protection and trustworthiness are considered right from the beginning, during the algorithm’s design, opening completely new possibilities in the age of digitization”, explains data expert Andreas Trügler of Know-Center.

As part of the COMET module “DDAI – Explainable, Verifiable and Privacy-Preserving Data-Driven Artificial Intelligence”, which is funded by the research company FFG, the state of Styria and the Federal Ministry, Trügler and his team are developing new and secure algorithms and methods to evaluate confidential and sensitive data. Top priority being not to compromise privacy and data protection. The module covers all parts of the data processing pipeline: from data infrastructure, cryptographic, algorithms and machine learning methods for secure data processing to explainable AI. So called “black box models” must not be used as AI algorithms. That means processes within complex neural networks must be transparent and AI’s results explainable and comprehensible. Overall, the goal is to get a better basis for decision-making that paves the way for secure and trustworthy AI solutions.

Diverse application potential

The research project’s results are integrated directly in the participating company partner’s industrial requirements and are tested within a particular application environment. “In addition, we have started to extend our knowledge to quantum algorithms and quantum computers,” Trügler states.

In fact, the new technologies can be applied in many ways: companies could gain competitive advantages by evaluating data together with their suppliers or competitors without being exposed to the risk of disclosing trade secrets or open doors to industrial espionage. For example, supply chains can be analyzed, and production data evaluated using encrypted machine learning without revealing their content. That way, you achieve better results benefiting all parties involved.

In medical research, often-sparse data sources for rare diseases could be evaluated across national borders jointly and encrypted, or the spread of infectious diseases analyzed using simulated data.

Further information: