UNIGE document Chapitre d'actes
previous document  unige:47676  next document
add to browser collection

Privacy-Preserving Multimodal Person and Object Identification

Published in Proceedings of the 10th ACM Workshop on Multimedia & Security. Oxford (UK) - Sep 22-23 - ACM. 2008, p. 177-184
Abstract In this paper we investigate multimodal biometric person and object identification. We model this process of multimodal identification as multimodal multiple hypothesis testing with independent modalities. We analyze theoretical performance limits that can be attained in such a multimodal protocol in terms of exponents of average error probability. Furthermore, we address a privacy related issues in this paper. In particular, we consider performance/privacy trade-off due to the indirect multimodal identification performed in a secret subspace and approximate the obtained performance limits using properties of random projections. Finally, a set of experiments exemplifies our findings.
Keywords WatermarkingMultimodal identificationrandom projections, Chernoff dis- Tanceprobability of error
Full text
Research groups Computer Vision and Multimedia Laboratory
Stochastic Information Processing Group
Multimodal Interaction Group
(ISO format)
KOVAL, Oleksiy, VOLOSHYNOVSKYY, Svyatoslav, PUN, Thierry. Privacy-Preserving Multimodal Person and Object Identification. In: Proceedings of the 10th ACM Workshop on Multimedia & Security. Oxford (UK). [s.l.] : ACM, 2008. p. 177-184. doi: 10.1145/1411328.1411359 https://archive-ouverte.unige.ch/unige:47676

307 hits

0 download


Deposited on : 2015-03-06

Export document
Format :
Citation style :