CISPA Faculty Dr. Franziska Boenisch first came into contact with machine learning during her IT studies. "I was hooked straight away," she says enthusiastically in the podcast. Above all, she wanted to understand what happens within machine learning models. "If we don't know how these models work, we can't understand how the data comes out of these models by mistake or at the wrong moment," she explains. As an example of a practical application, Boenisch mentions the task of a language model to create a CV from the personal data entered. She wants to build models in such a way that the data can only be used for the job, but cannot be passed on. The technical challenge here is not only to protect privacy, but also to allow the models to make good predictions at the same time, explains Boenisch in an interview.
Open source models and the role of politics
Boenisch's research is made more difficult by the fact that the models are concentrated in the hands of just a few companies. "We have to pay Open AI to be allowed to do research," says Boenisch in the podcast. Open source models would be an alternative in her opinion. For her, it is crucial that research always keeps pace with the state of the art, which, however, requires large financial resources. And this also requires support from politicians, which is why knowledge transfer is an important issue for her: "We need more experts who can bring the output from science back into politics." Unfortunately, there is often not enough time for knowledge transfer in everyday life. "It would be important to provide scientists with better financial resources so that they don't just have to worry about publishing and acquiring new funding, but have funded blocks of time where they can dedicate themselves to knowledge transfer," says Boenisch in the interview.
Early promotion of young female researchers is important
During her studies, Boenisch wished she had had a female mentor. "I didn't have a role model to look up to," she explains in the podcast. "Mentoring programs are important so that we don't lose the few women in our field," she says with conviction. In the SPRINT ML Lab, which she founded at CISPA, she learned that simply advertising a position and hoping that women will apply is not enough. She herself has had good experiences with approaching female students during their bachelor's degree and introducing them to research by working as a student assistant, for example. But she is convinced: "There needs to be a big mindshift on the part of the supervisors." Because while men often emphasize that they can do everything, women tend to point out that they still have to work things out, she says in the podcast. This requires a rethink in application processes. Boenisch has a clear goal in mind: "I want to attract more female doctoral students to my group".
TL;DR, short for "Too Long Didn't Read", is CISPA's podcast, with "Women in Cybersecurity" as a special edition. We have been on air since 2022 and are available on all common podcast platforms. Every month, we talk to CISPA researchers about their work on cybersecurity topics and artificial intelligence and try to ask them exactly the questions that listeners are asking themselves. Our aim is to explain complex topics in simple language. Since people from 49 nations work at CISPA, the conversations are sometimes conducted in German, sometimes in English.
