TechEdge AI Interview with Arnd Baranowski, CEO, Oculeus

With a successful track record of 30 years in technical and software development, please share your journey and lessons learned along the way?

Our journey at Oculeus has certainly been an exciting one. We started out as a consulting firm serving telecoms operators in our local market. With time, we developed a strategic portfolio of software-based solutions and have grown into an international company with customers in a range of vertical markets around the world.

One of the prominent lessons that we have learned during our journey is that customer requirements never stand still. As providers of technology, we always need to be forward-thinking and anticipate future requirements.

We are proud of the strong social value of our technology for protecting individuals and organizations against the ongoing threats of telecoms fraud. The increasing digitization of daily life has increasingly moved the attack-surface of cybercriminals into the telecommunications domain. There are constantly new and innovative fraud methods being developed by cybercriminals, while many ongoing fraud scenarios stubbornly persist simple because they continue to work.

Moving forward, one of the biggest challenges concerns the abilities of AI to equip fraudsters with improved techniques. The only answer to this is to fight fire with fire, which is the reason why have embedded AI into our technology and solutions.

Telecom fraudsters are constantly coming up with new schemes. How can AI be used to stay ahead of these evolving threats and identify previously unseen patterns of fraud?

AI excels at recognizing very subtle patterns from within very large data sets. This is no secret. Telecoms fraud attacks move far too quickly to rely on lengthy human analysis. This is why AI is essential for seeing the unseen and efficiently positioning providers of telecommunications services to better combat telecoms fraud.

By using AI on the metadata of telecom calls, it is possible to identify emerging patterns of new fraudulent behavior that may involve a variety of different parties – not only subscribers or B2B customers, but also business partners, including the network of interconnect and wholesale providers and suppliers who ship call traffic from A to B to C.

Can AI automate some aspects of the fraud prevention process, such as flagging suspicious activity or even taking initial steps to block fraudulent calls?

AI can automate the early detection of emerging and new fraudulent patterns in telecommunications traffic and activities. We see that AI can be very useful in acting as an early warning system. Alternatively, without AI in place, we would need to allow time for attacks to ‘mature’ to the point where they form a pattern recognizable to either a human or a traditional rules-based anti-fraud system. Today, this alternative is no longer sufficient and is allowing telecoms fraud to cause significant financial damage.

AI clearly excels at detecting fraud events that otherwise would be missed. AI can be further used to automatically assign a risk-level, which in turn can define whether there is an existing threat that will likely require blocking.

While AI is powerful, human expertise is still crucial. How can AI and human analysts work together most effectively to combat telecoms fraud?

“Computers are incredibly fast, accurate, and stupid; humans are incredibly slow, inaccurate, and smart; together they are powerful beyond imagination”.

The quote is often attributed to Albert Einstein, although there is some debate about who actually said this. Regardless of who this statement should be correctly attributed to, it is incredibly wise and highly applicable to the use of AI.

AI can detect emerging patterns early on and recommend actions based on predefined targets or thresholds. However, it is human expertise that makes sense of these outputs to make decisions accordingly.

There are two sides of AI with human interactions that make a full circle. First, reviewing output from AI to decide which actions should be taken and, second, providing feedback to AI models for future evaluations to make more accurate evaluations.

In parallel, emerging trends in AI Governance, such as the current draft of the “EU AI Act”, will give citizens the right to challenge decisions or actions taken by AI systems. This will impose on organizations an obligation to explain “how and why they took that action”. Simply explaining that “the AI did it” will no longer meet the requirements of the law.

Are there any real-world examples of telecom companies successfully using AI to combat fraud? What can we learn from their experiences?

In our telecoms fraud prevention offering, we have developed a multi-prong AI module that evaluates traffic on three levels: (1) general traffic behavior to identify changes, (2) call-level evaluations assessing the fraud risk based on similar or previously identified pattern and (3) incident level evaluations where cases that have not reached the defined threshold, but closely follow fraudulent patterns are marked for further review.

Customers using our software report that more fraud cases are successfully identified and blocked in less time, thanks to AI technologies.

Furthermore, our application of AI learns from its own experiences. What this means is whenever a fraud case is confirmed, then our AI is trained on this case data to further refine the effectiveness of the detection engine.