%0 Conference Proceedings
%T Feature Selection by User Specific Feature Mask on a Biometric Hash Algorithm for Dynamic Handwriting
%+ Brandenburg University of Applied Sciences
%A Kümmel, Karl
%A Scheidat, Tobias
%A Arndt, Christian
%A Vielhauer, Claus
%Z Part 1: Research Papers
%< avec comité de lecture
%( Lecture Notes in Computer Science
%B 12th Communications and Multimedia Security (CMS)
%C Ghent, Belgium
%Y Bart Decker
%Y Jorn Lapon
%Y Vincent Naessens
%Y Andreas Uhl
%I Springer
%3 Communications and Multimedia Security
%V LNCS-7025
%P 85-93
%8 2011-10-19
%D 2011
%R 10.1007/978-3-642-24712-5_7
%K biometrics
%K dynamic handwriting
%K biometric hashing
%K user bitmask
%K feature selection
%Z Computer Science [cs]
%Z Computer Science [cs]/Networking and Internet Architecture [cs.NI]Conference papers
%X One of the most important requirements on a biometric verification system, beside others (e.g. biometric template protection), is a high user authentication performance. During the last years a lot of research is done in different domains to improve user authentication performance. In this work we suggest a user specific feature mask vector MV applied on a biometric hash algorithm for dynamic handwriting to improve user authentication and hash generation performance. MV is generated using an additional set of reference data in order to select/deselect certain features used during the verification process. Therefore, this method is considered as a simple feature selection strategy and is applied for every user within the system. In our first experiments we evaluate 5850 raw data samples captured from 39 users for five different semantics. Semantics are alternative written content to conceal the real identity of a user. First results show a noticeable decrease of the equal error rate by approximately three percentage points for each semantic. Lowest equal error rate (5.77%) is achieved by semantic symbol. In the context of biometric hash generation, the reproduction rates (RR) increases by an average of approx. 26%, whereas the highest RR (88.46%) is obtained by semantic symbol along with a collision rate (CR) of 5.11%. The minimal amount of selected features during the evaluation is 81 and the maximum amount is 131 (all available features).
%G English
%Z TC 6
%Z TC 11
%2 https://inria.hal.science/hal-01596201/document
%2 https://inria.hal.science/hal-01596201/file/978-3-642-24712-5_7_Chapter.pdf
%L hal-01596201
%U https://inria.hal.science/hal-01596201
%~ IFIP-LNCS
%~ IFIP
%~ IFIP-TC
%~ IFIP-TC11
%~ IFIP-TC6
%~ IFIP-CMS
%~ IFIP-LNCS-7025