The Dutch company Flytxt develops software products for marketing automation with machine learning and artificial intelligence. It does this for over one hundred companies in fifty countries. With the introduction of the European privacy legislation (GDPR) in 2018, all Flytxt software products have had to comply with European requirements aimed at protecting commercial and customer or patient sensitive information. Flytxt has been working with TNO since 2015 to ensure that it can meet these demands.
Want to know more?
Want to know more about secure multi-party computing and its possible applications? Then contact Thijs Veugen.
Together with its partners, such as Flytxt, TNO develops algorithms for encrypting privacy-sensitive data, using the technology of secure multi-party computation (MPC) to this end. "This is a safe way for parties to undertake joint computation with their own secure database," says Thijs Veugen, a cryptographer and senior scientist at TNO. "MPC enables you to analyse shared data in a virtual way without actually sharing or exchanging data. The participants jointly determine who gets to see the final result of the desired computation."
"TNO has a lot of scientific knowledge in-house and can also demonstrate the practical applications of that knowledge"
"Well before the introduction of European Privacy legislation, we were already working on making our software privacy friendly," says CTO Prateek Kapadia of Flytxt. "We asked TNO to help us with this. A logical choice because TNO has an enormous amount of scientific knowledge in house and can also demonstrate many practical applications of that knowledge. My experience is that Thijs Veugen and his colleagues also speak our language and really understand our business challenges. TNO is a research partner you can count on.”
From its Dutch base Flytxt wants to strengthen its position as a privacy-friendly marketing technology for business-to-consumer or “B2C” companies globally. B2C companies, like financial institutions, retail banks, car manufacturers, airlines and large telecom companies work directly for end consumers.
Flytxt’s marketing technology includes recommendation systems for 'telcos', whereby all kinds of private data shared by the consumer with the telco must be used securely: your consumption behaviour, which products you use, how you use them, and so on. The analysis and recommendation software developed by Flytxt enables a company to tailor its offer to a customer, something that is extremely difficult to achieve with strict privacy legislation. However, MPC can offer a solution to this.
In close cooperation with Flytxt, Veugen has developed three patents based on MPC technology. These patents have been or will be incorporated in the Flytxt software products. The first patent concerns a kind of automatic stop incorporated in the Flytxt software products whenever the threat arises of computation making use of the secure data of too few people. In principle, this kind of computation makes it possible to trace the encrypted personal data back to specific persons. And that is prohibited by European Privacy Law.
2 plus 3 is 5
Veugen: "You can look at it like this: 2 plus 3 is 5. But if you know that the conclusion is 5, you can deduce something from it: like 1 plus 4 or 2 plus 3. So the conclusion says something about the data. And we must prevent that from happening. This can be done by working with a sufficiently large minimum number of people. So, just to give you a figure, about a hundred or a thousand people."
"Nobody - not even the server - knows which persons have been selected for the computation"
The second patent that can be incorporated in the Flytxt software builds on the first patent but provides even better privacy protection. The personal data for a computation are selected at random. Nobody - not even the server - knows which persons have been selected for the computation.
The customer population of a telecom company can be clustered into, for example, the frequent callers, the prepaid callers, etc. You can apply an algorithm to your own clusters and draw conclusions based on them. But those conclusions will become more accurate and better if the computation is undertaken jointly with more telcos. The third patent that Veugen has developed for Flytxt relates to so-called hierarchical cluster methods.
"Telecommunications companies working with this Flytxt software can securely start doing joint computation"
"These methods already exist," says Veugen, "but what has been lacking so far is a privacy-friendly solution to them. I came up with an MPC solution that you can apply to the virtual shared database of telecom companies. After the computation, they only know what their own clusters look like. This allows any company to make recommendations to their own customers." This patent is being incorporated in the Flytxt software so that the telecom companies working with this software can do joint computation securely.
Together with a few other people from TNO, Flytxt is also exploring other domains than just telecom companies for the use of their software. Kapadia: "We are now exploring logistics and mobility as a potential new market. TNO is a dependable party that has many contacts itself. And we are keen to make use of that!”