A training algorithm for optimal margin classifiers
COLT '92 Proceedings of the fifth annual workshop on Computational learning theory
Machine Learning
On-Line Fingerprint Verification
IEEE Transactions on Pattern Analysis and Machine Intelligence
Biometrics, Personal Identification in Networked Society: Personal Identification in Networked Society
Training Support Vector Machines: an Application to Face Detection
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
Handbook of Fingerprint Recognition
Handbook of Fingerprint Recognition
Cryptographic Key Generation from Voice
SP '01 Proceedings of the 2001 IEEE Symposium on Security and Privacy
Support Vector Machines: Training and Applications
Support Vector Machines: Training and Applications
Secure smartcardbased fingerprint authentication
WBMA '03 Proceedings of the 2003 ACM SIGMM workshop on Biometrics methods and applications
Filterbank-based fingerprint matching
IEEE Transactions on Image Processing
Random number generation based on fingerprints
WISTP'10 Proceedings of the 4th IFIP WG 11.2 international conference on Information Security Theory and Practices: security and Privacy of Pervasive Systems and Smart Devices
| Hi-index | 0.00 |
We show an architecture to automatically generate cryptographic keys using the FingerCode as defined by Jain et al. [4]. The FingerCode is obtained from gray scale fingerprint images. The architecture uses a classifier to compensate for the natural variability on the FingerCodes. In a training step the FingerCodes of the fingerprint samples for registered users are obtained; then random binary codes are assigned to each set of FingerCodes from the same finger, and finally an array of Support Vector Machines (SVM) is trained to associate the FingerCodes to their assigned random binary key. Each SVM is independent and assigns one bit, allowing the construction of binary keys of arbitrary length by adding and training more SVMs. To test the system, different set of fingerprint images from the same fingers used on the training step were used. The FingerCodes were calculated used as input to the SVM array to generate the assigned keys. Experimental results obtained using fingerprints selected from the FVC2000 and FVC2002 databases show results up to 90% performance on generating valid keys.