Growing numbers of companies are investing in machine learning and artificial intelligence to optimize their internal work processes and offer customers new services by means of recommender systems. “New tools of this type are often rapidly designed and implemented in quick succession. Unfortunately, this tends to mean that security-related issues fall by the wayside,” says statistician Verena Battis from Fraunhofer SIT. She studies the risks that machine learning systems pose to data privacy as part of ATHENE’s Privacy Risks and Safety in Machine Learning (PRisMa) project. Battis says that data theft and manipulated algorithms or data are among the negative outcomes of risky practices. As computer scientist Franziska Boenisch from Fraunhofer AISEC explains, it’s best to err on the side of caution when it comes to data protection: “User data is frequently used to train machine learning or artificial intelligence models. However, machine learning practitioners may lack an understanding of how individual data points can influence a model and whether these can be extracted from a trained model by cyber attackers.” The aim of the research project is to gain valuable insights on the extent to which security and privacy are factored into and implemented in machine learning in everyday life.

You can take part in the online survey if you use machine learning processes in your professional or personal life. It only takes 15 minutes to complete.

https://websites.fraunhofer.de/ML_security/index.php/149369?Start1=A3

The photo is available for download by the media. It is free of charge if used in the context of the press release and due credit is given to the photographer.