Fingerprint Single-image Morph Attack Detection
The aim of fingerprint single-image morph attack detection is to assess whether a suspected fingerprint is a morph (also called a double-identity fingerprint) or not.
Benchmarks
Currently, a benchmark is being used to evaluate the performance of fingerprint single-image morph attack detection approaches:
- Fi-SMAD-1.0: A double-identity fingerprint benchmark created from the FVC2002 DB1-A database. It contains 800 real identity fingerprints from 100 fingers, 8 impressions per finger and three sets of 100 double-identity fingerprints produced using three different generation approaches (Im, En, and IITBBS) [1] [2] [3].
Performance Evaluation
For each algorithm, bona fide (evaluating bona fide fingerprints) and morph (evaluating double-identity fingerprints) attempts are performed to compute Bona fide Sample Classification Error Rate (BSCER) and Morphing Attack Classification Error Rate (MACER). As defined in [4] the BSCER is the percentage of bona fide samples incorrectly classified as morphed samples while the MACER is the proportion of morphing attack samples incorrectly classified as bona fide samples.
The following performance indicators are computed:
- EER (detection Equal-Error-Rate: the error rate for which both BSCER and MACER are identical)
- BSCER10 (the lowest BSCER for MACER≤10%)
- BSCER20 (the lowest BSCER for MACER≤5%)
- BSCER100 (the lowest BSCER for MACER≤1%)
To calculate the above indicators, both bona fide and double-identity fingerprints are provided as input to the algorithm being evaluated.
Results
In this section, the results obtained on the Fi-DMAD-1.0 benchmark by two well-known CNNs (i.e., AlexNet [5] and InceptionV3 [6]) specifically trained for the fingerprint single-image morph attack detection task are presented.
The following table report the results obtained on the Im double-identity set.
| Method |
EER |
BSCER10 |
BSCER20 |
BSCER100 |
| AlexNet |
0.02% |
0.00% |
0.00% |
0.00% |
| InceptionV3 |
0.00% |
0.00% |
0.00% |
0.00% |
The following table report the results obtained on the En double-identity set.
| Method |
EER |
BSCER10 |
BSCER20 |
BSCER100 |
| AlexNet |
0.02% |
0.00% |
0.00% |
0.04% |
| InceptionV3 |
0.00% |
0.00% |
0.00% |
0.00% |
The following table report the results obtained on the IITBBS double-identity set.
| Method |
EER |
BSCER10 |
BSCER20 |
BSCER100 |
| AlexNet |
0.21% |
0.00% |
0.00% |
0.04% |
| InceptionV3 |
45.73% |
100.00% |
100.00% |
100.00% |
Terms and Conditions
All publications and works that cite the information reported on this web page must reference sources [1], [2], and [7].
Bibliography
[1] M. Ferrara, R. Cappelli and D. Maltoni, "On the Feasibility of Creating Double-Identity Fingerprints", IEEE Transactions on Information Forensics and Security, vol. 12, no. 4, pp. 892-900, April 2017.
[2] M. Ferrara and A. Franco, "Deliverable 10.3 - Morphing techniques for secondary biometric characteristics", iMARS European Project, September 2023.
[3] I. Goel, N. B. Puhan and B. Mandal, "Deep Convolutional Neural Network for Double-Identity Fingerprint Detection", IEEE Sensors Letters, vol. 4, no. 5, pp. 1-4, May 2020.
[4] ISO/IEC DIS 20059, "Methodologies to evaluate the resistance of biometric systems to morphing attacks", 2024.
[5] A. Krizhevsky, I. Sutskever and G. E. Hinton, "ImageNet Classification with Deep Convolutional Neural Networks", in Advances in Neural Information Processing Systems, Curran Associates, Inc., 2012.
[6] S. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens and Z. Wojna, "Rethinking the Inception Architecture for Computer Vision", arXiv, 2015.
[7] N. Gonçalves, "Deliverable 10.4 - Attack resolving solutions for finger, iris, and 3D face image manipulations", iMARS European Project, June 2024.
|