Main focus: Privacy in Machine Learning
Twitter handle: @fraboeni
Languages: German, English, French, Dutch, Spanish, Turkish
Topics: machine learning, artificial intelligence, privacy, trustworthy ai
Services: Talk, Consulting, Interview
Willing to travel for an event.
Willing to talk for nonprofit.
I am doing research on privacy in machine learning.
The privacy risks of machine learning (ML) have only been an issue for a few years, but they affect everyone whose data is used to train ML models. The risks involved are diverse: for example, it is possible to extract potentially sensitive training data from trained models. Thus, an attacker can infer the underlying data by accessing the model. Additionally, given an existing model, it is also possible to predict with relatively high accuracy whether a particular data point was used for training. Imagine we have a model that detects health problems in alcoholics. Such a model would need to be trained on data from alcoholics. If it can be found out that a certain individual was used for training, it can be inferred that he or she is an alcoholic, thus violating the privacy ferry.
In my research, I investigate the causes of such privacy problems and work on the development of methods to reduce the risks. For this purpose, I am intensively engaged in the topic of privacy, but also security and data protection.
Translated with www.DeepL.com/Translator (free version)
09/2019 - current: Researcher
Department of Secure Systems Engineering, Fraunhofer AISEC, Germany
Duties included: Working in industrial and academic research projects concerning private machine learning and data protection
Supervisor: Prof. Dr. Marian Margraf
09/2017 - 01/2019: Research Assistant
Dahlem Center for Machine Learning and Robotics, Freie University Berlin, Germany
Duties included: Programming tasks in MATLAB and Python, literature research for projects, support writing of scientific papers and funding applications
Supervisor: Prof. Dr. Tim Landgraf
08/2016 - 10/2016: Visiting Undergraduate with DAAD RISE scholarship
National Chung Cheng University, Chiayi, Taiwan
Duties included: Implementing neural networks for food image classification
Supervisor: Prof. Ph.D. Wei-Ta Chu
01/2016 - 07/2016: Working Student
Fraunhofer Institute FOKUS, Berlin, Germany
Duties included: Implementing scientific demonstrators for predictive maintenance with Apache Spark and Flink
Supervisor: Prof. Dr. Adrian Paschke
Examples of previous talks / appearances:
In diesem Meetup Talk stelle ich Methoden vor, um Machine Learning privatsphärebewahrender zu getstalten. (Leider gab es ein technisches Problem, daher läuft meine Kamera nicht mit.)This talk is in: English
Mit diesem Science Slam zum Thema Differential Privacy habe ich in Tomsk, Russland, beim deutsch-russischen Science Slam den 1. Platz gewonnen.This talk is in: German