 |
Gábor Gulyás, Ph.D.Owner, Project Manager
|
What’s in a face (imprint)?
State of the art facial recognition uses deep learning, where neural networks produce embeddings (e.g. a 128d vector of float values). These then can be used to identify individuals by comparing the similarity between two embeddings. But can a single embedding tell enough to identify its original subject?
This talk is about understanding the privacy of facial recognition embeddings. What details are stored in them and how? Do they pose privacy risks? (Hint: sure they do!) If yes, what can we do to mitigate these issues? Can we somehow just remove personal data while also preserving utility?
BIO
Gábor is a passionate expert in the field of data protection, with over 15 years of experience in developing technical solutions, accumulating valuable expertise both in research and business. Currently, Gábor serves as a team leader at Vitarex, where he spearheads cutting-edge machine learning projects. One of his notable focuses is the development of a privacy-preserving facial recognition solution.
Prior, Gábor made significant contributions to SEON as a researcher (since 2017), swiftly progressing to the position of tech lead for the device fingerprinting team. Before that, he worked as a researcher at Budapest University of Technology and Economics and as a PostDoc at Inria (France).
Throughout his career, Gábor has been actively involved in numerous data protection projects, focusing on data de-anonymization, device fingerprinting and privacy-enhancing technologies. His work has been recognized on a broader scale, with his findings being presented to esteemed bodies such as the European Parliament.