Forensic Machine Learning Technology Center
Founded in 2022, the Forensic Machine Learning Technology Center (ForMaLTeC) focuses mainly on basic research for machine learning as well as application-oriented methods using machine learning and deep learning in forensic science. It is the internal core facility for both the Forensic Institute in Zurich (FOR) and the Zurich Institute for Forensic Medicine (IRM) for all questions related to statistics and mathematical modeling.
Wherever forensic questions need to be identified and evaluated from data, ForMaLTeC offers a support. This support can range from data collection to the evaluation of already existing data with the help of computer algorithms and statistical models.
We support the expert in such a way that large amounts of data from one or even several sources can be evaluated more efficiently and effectively at the same time. This is particularly helpful for comparative studies and individualizations (same source vs. different source). For instance in the case of individual comparisons, or the comparison of an object that is part of a trail, or the comparison of traces with existing databases.
The center is co-directed by Dr. Michael Bovens (FOR) and PD Dr. Akos Dobay (IRM).
Dr. Michael Bovens (FOR)
"Machine Learning" (ML) and "Deep Learning", so-called Deep Learning (DL), are among the latest scientific research areas. This development came about as a result of the constant evolution of microprocessors (CPUs) and, above all, graphics cards (GPUs), which meanwhile not only control screens, but are also used as computing units. The primary goal of ML/DL is to automate parts of the work process and basic decision-making more objectively.
Documents, Internet trades, etc. are increasingly being forged and misused with the trend towards digitalisation. Therefore, the development of new methods and the development of basic expertise in this area for early detection or case clarification are increasingly in demand.
Mathematical modeling is a tool. Much more important are the data selection, the experimental design, and the validation. For instance, do the data represent a real case scenario? How are the data distributed? Equally central is a good knowledge of the model limitations. Ultimately, mathematical modeling includes checking the results with the case conditions such that the interpretation of the results remains within the expertise of a forensic expert.