Research is typically (mis)understood as a boring, tedious, long-term task decoupled of any real-world impact. This is not the case for SnT, who carries out highly relevant, internationally competitive research in information and communication technology (ICT) with socioeconomic impact. They frequently engage in demand-driven, collaborative projects with their partners – of which they have nearly 60 – both in Luxembourg and further afield, to progress the state-of-the-art. Their Partnership Programme has been a mainstay in their operations over the last decade and has helped them address real challenges that confront private companies and public sector bodies today.
57 companies are in the Partnership Programme.
413 employees are working at SnT.
8 laboratories are available for the researchers.
€181 million from external funding invested in the centre.
Alignment with Local Market
SnT’s strategic research areas fall in line with the Luxembourg economy’s push for digitalisation. Their research focuses on secure and compliant data management and FinTech, cybersecurity, space systems, autonomous vehicles and the internet of things. Their projects intend to create a real, durable and competitive advantage to companies based in Luxembourg or beyond.
With work spanning from start-ups to household names, one example of their partners is financial giant PayPal . While projects are currently ongoing with PayPal, SnT also holds a research chair – backing from PayPal and the Fonds National de la Recherche (FNR) to conduct research into the future of financial services. Prof. Gilbert Fridgen holds this prestigious position, and has built an international team of researchers with a diverse set of expertise. The FINATRAX research group features a wide range of disciplines, including computer scientists, engineers, political scientists, economists and psychologists. Together, they take a 360-degree approach to solving industry challenges using technologies from AI to blockchain to find reliable and secure business solutions.
Headed by Björn Ottersten , SnT has a unique model to foster collaboration between applied research and industry, with a direct and immediate link to everyday challenges businesses face. Since their launch in 2009, they have undergone rapid development that has seen them recruit top scientists, launch over 90 European Union and European Space Agency projects, create a Technology Transfer Office (TTO) to protect and license intellectual property (IP), launch five spin-off companies and create a dynamic research environment with 400+ people.
These endeavours have reinforced our impact in Luxembourg and increased our international visibility.
Their penchant for developing spin-offs will see them soon launch another, DigitalUs – a company with their sights set on providing financial firms with an automated answer to time-consuming know-your-customer and anti-money laundering checks. The unique technology leverages “A wealth of data that is publicly available in order to produce a unified digital footprint,” according to Beltran Fiz , CEO of DigitalUs.
One of SnT’s most significant departments, the TTO, brings together researchers with industry. “Transferring knowledge and technology developed in the University to public or private actors, as well as the general public, is an important goal of SnT,” said Dr. Carlo Duprel , head of the TTO, as he explained the importance of interacting with the ‘outside’ world. “The way we do this is threefold. Firstly, the outcome of a typical project is a prototype demonstrating a technology can solve a business challenge. Secondly, the partner’s staff working with SnT researchers develop their competences in, and mastery of, the technology. Lastly, we also train people. We recruit students globally, train them in industry projects and then release them to be potentially hired by the partner.”
“Ensuring that research isn’t stuck in scientific publications and is positively impacting businesses in Luxembourg, and the rest of the world, is an important function of SnT. The centre enables us to give support to companies and public bodies who want to invest in innovation,” Duprel said.
For companies partnering with SnT, the investment is split evenly either side. But an important benefit they often see is government funding. As a research institution, SnT helps to attract government funding through its various financial instruments that aim to foster research and development in companies.
“We, as a centre, know how to construct projects in line with the government’s objectives and are eligible to receive a level of co-financing. This encourages our partners to invest in research, as the government shares the inherent risk associated with innovation projects.”
Helping the financial sector to implement artificial intelligence
One of the most significant technological advancements of our time has been the emergence of artificial intelligence. This technology, which allows for time-consuming tasks to be automated, is highly relevant to financial institutions and FinTech players. Prof. Jacques Klein, a chief scientist and head of the TruX research group at the Interdisciplinary Centre for Security, Reliability and Trust (SnT), has done extensive work on building trustworthy software for this sector.
How are you collaborating with businesses in Luxembourg’s financial sector?
We currently work with many institutions across Luxembourg’s financial landscape, including BGL BNP Paribas . My team and I explore explainable AI techniques to bring insight into the internal decision-making process of machine learning algorithms. It’s important for bank experts to be able to understand the decisions that an AI makes, to comply with any regulations. With BGL, we are exploring the state-of-the-art explainability algorithms and developing new machine learning algorithms, with the goal of making sure any AI decision can be fully explained and understood by banking experts.
Why explainable machine learning over another type of AI?
The reason we are developing explainable machine learning is because we want to understand how an AI makes a decision. It would be like showing your work on a mathematical equation. In my work, I try to figure out what important data are contributing to a decision, either positively or negatively. Different types of AI exist – for example, deep learning. But by design, it’s very difficult to interpret this type of reasoning. The type of AI we implement will depend on the partner’s challenge and the type of regulations they face in their industry.
What’s been your experience in investigating new AI techniques for partners like BGL?
Collaborating with researchers from SnT is highly beneficial for BGL, as the researchers explore state-of-the-art explainability algorithms. This domain is relatively new, hence SnT and BGL are strongly committed to research to become a leader in explainable AI for financial services. Constant interactions between highly skilled experts, Ph.D. students and product owners of BGL provide new ideas, leading to publishing innovating papers and industrial solutions.
Why security is vital for robotics systems
When we think of cybersecurity, robotics systems may not immediately come to mind, but it makes sense that they need protection against hackers. We spoke to Prof. Radu State , head of SnT’s SEDAN research group, who specialises in cybersecurity. His group includes Dr. Antonio Ken Iannillo , a research scientist, and Dr. Sean Rivera , a postdoctoral researcher. Together, they are part of the EU’s CONCORDIA project, developing innovative cybersecurity solutions in partnership with Luxembourg’s Direction de la Défense .
What are the current issues you tackle in your work on security and robotics?
Some robotics systems operating today don’t have proper security mechanisms in place. This means robots are being deployed, but the software that controls them and their security systems may be considered as an afterthought. Sometimes, they are responsible for critical missions. So, for the safety of those operating them or anybody around them, they need security. Unfortunately, many robotics systems are designed to be very efficient at machine learning, in being able to drive and make decisions, but their security systems are left vulnerable to hacking. Anything that can make them unsafe to humans needs to be fixed as a priority.
How does your work address this situation?
We looked at the most used robot operating system (ROS). It’s open source, meaning it’s available to anyone. We extended that system with security mechanisms, implementing a kind of firewall that can protect the robot. We also implemented a tool that allows you to assess the security of a robot and tell you whether it’s vulnerable or improperly configured.
Is it really as simple as putting in a firewall?
Well, that depends. When you consider security, you have to think about defining the threat landscape. For example, what are the potential cyber-threats, and how can they be identified? Then, you have to think about what the mitigation plan will be and then, you start filling in the gaps, like putting a firewall in place. You have to think about the impact of the overhead that you’re bringing into the system and you have to think about fault tolerance.
A global first for space communications research in Luxembourg
Luxembourg is quickly realising its ambition to develop a private space industry. A key partner in this endeavour has been SnT’s SIGCOM research group, who specialise in wireless and satellite communications, networking and radar systems. Prof. Symeon Chatzinotas is co-head of the group, who works with industrial partners on space-based 5G communications.
What role does SnT have in Luxembourg’s space industry?
Our main role is to support long-term collaborative research with the space industry. While our partnerships started traditionally with established companies like SES , now we are beginning to engage more with start-ups and smaller companies, such as Lift Me Off .
How does your work with 5G communications apply to the space industry?
Here, at SnT, we have one of the largest satellite communications teams. We’re very active in 5G space communications, and pioneers in this area. We are lucky to benefit from extensive experimentation facilities in-house, but also access to real satellites through SES. We have already successfully tested 5G technologies over satellites, and we are involved in the groundbreaking European Space Agency project to verify and validate a full-stack 5G transmission over a geostationary satellite in the next year – it’s going to be a worldwide first!
What kind of projects are you currently working on?
Our project entitled 5G SPACE focuses on two main scenarios. One is mega satellite constellations scenario, such as those that private companies like SpaceX are launching. We will be able to simulate advanced low-earth orbit (LEO) satellite constellations in the lab. It will be a unique facility because right now, these are only just being deployed into orbit, so our knowledge is limited. The second scenario is looking to establish communications on the Moon’s surface.
Another project, SPAICE, is examining how artificial intelligence can be used to improve satellite communications by putting an artificial intelligence chipset into a CubeSat, a satellite measuring just 10cm on each side.
Driving the future of self-driving cars in Luxembourg
Self-driving cars look to be the future of the automotive industry. So when innovators at SnT began an autonomous vehicle project, building one of their own was a natural step. Dr. Georgios Varisteas , a research associate at SnT’s Services and Data Management (SEDAN) research group, was the lead engineer in creating ‘Junior’, the heart of the 360Lab .
What was the goal of the autonomous vehicle project?
Usually, our research projects at SnT start with a defined budget and industrial partner. For us, Junior was more of a passion project. Our initial goal was to reproduce a paper published by Nvidia Corporation , the US-based computer graphics company at the forefront of AI development. They implemented ‘imitation learning’ or ‘end-to-end learning’, a large-scale AI program that sees its environment and decides how to drive. In autumn 2018, after just six months, we demonstrated a functioning autonomous car using imitation learning to smoothly traverse a path marked by traffic cones.
What is it like to put an AI system into a car?
Putting AI into a car is worlds away from putting it in a computer. If we put AI into a computer, the consequences of its prediction being wrong are not serious. However, if a car makes the wrong decision, it can endanger lives. For that reason, we have to take a lot of care, implement safety measures and set in place certain guarantees.
Where is the project today?
Since 2018, we developed a unique software platform called ‘Robocar’ that is capable of powering a high-performance autonomous vehicle but using normal hardware. The software is currently being tested in Junior under various weather conditions, which are currently a challenge even for the state-of-the-art autonomous vehicles that exist at the moment.