Workgroup: "Differentiable Privacy"


Head of group: Dr. Martin Kroll

Members:

  • Prof. Dr. Holger Dette
  • Martin Dunsche
  • Carina Graw
  • Dr. Tim Kutta


  • Description:

    In statistics, the use of sensitive data often results in the risk that the privacy of individuals can be endangered from statistical analyses. Typical approaches, such as anonymization, have lost their importance due to the rapid development of artificial intelligence, and a new method of privatization, called differential privacy, has been established. Conceptually, differential privacy guarantees every individual that he or she cannot be identified as a participant in a data analysis (e.g. statistics, machine learning). The basis of success is the mathematical formalization and further development of this concept. The workgroup "Differential Privacy" is interested in whether previous statistical methods can still be used in the context of differential privacy. One example is the interaction between privacy and statistical accuracy.