Our working world is becoming ever more digital, and it is our mission to embrace the opportunities opened up by new technologies. Because of our global reach and different divisions, there is an immense variety of projects around data at Bertelsmann. Find out more about the importance of data science in the context of an international media, services and education company.
Why is data important for your division?
In times of growing cyber crime and fraudulent online activities, it is important to take effective precautions against unauthorised access and fraudsters who are obtaining goods and services under false pretences. Preventing fraudulent cases starts with recognising suspicious activity. Analysing multiple data sources like access points, digital identity or online activity helps us identify different types of fraud risks.
How is data being applied in your division?
We identify legitimate and fraudulent events in online business by analysing the data of the device and its browser from the very start. By recognising trusted devices and defined transaction patterns, it is possible to automatically detect suspicious actions and to fight fraud at an early stage. Browser fingerprinting and behavioural biometrics allow detecting identity theft, synthetic identity, payment fraud and account takeover. Behavioural biometrics analyse more than 100 physical behavioural parameters, such as a user’s finger pressure on a mobile device, fingertip size, touch coordinates and device movement, in order to generate a unique profile for each user. This also allows us to distinguish between human and automated fraudulent activities.
What role does data play in your job?
As an end-to-end machine learning scientist, I’m involved in every step of a data project. I work closely with fraud experts and the business unit to develop ML-based solutions for our customers. As fraudsters are constantly evolving, we need to develop new solutions to prevent fraud. Together with other data scientists and developers, I take care of data collection, data preprocessing, models training and evaluation. After that, I deploy the models and run proofs-of-concept with the customers. This process is iterative, which means we go back and forth in testing ideas, including features, adding customer requests and deploying new models.