Amilab IJCB 2011 Poster
0 likes575 views
The document summarizes a multi-clue approach for detecting photo-based face spoofing attacks in face recognition systems. It fuses analysis of both static visual characteristics and video clues, such as motion and eye blinking. For static analysis, it extracts several visual representations from frames to compute scores. Video analysis examines motion and blinks. The scores are fused using different combination methods depending on the level of detected motion. Experimental results on a standard spoofing database show the fused approach is more effective and robust than static analysis alone, especially for higher quality spoofing attacks.
![Fusion of multiple clues for photo-attack detection in face
recognition systems
Roberto Tronci , Daniele Muntoni , Gianluca Fadda , Maurizio Pili , Nicola Sirena , Marco Ristori ,
1,2 1,2 1 1 1 1
Gabriele Murgia , Fabio Roli
1 2
1
Ambient Intelligence Lab, Sardinia DistrICT, Sardegna Ricerche, ITALY
2
DIEE, Dept. Electric and Electronic Engineering, University of Cagliari, ITALY
{roberto.tronci, muntoni, fadda, maurizio.pili, sirena, gabriele.murgia, ristori}@sardegnaricerche.it
{roberto.tronci, daniele.muntoni, roli}diee.unica.it
Introduction
SARDEGNA
RICERCHE We faced the problem of detecting 2-D face spoofing attacks performed by
placing a printed photo of a real user in front of the camera.
For this type of attack it is not possible to relay just on the face
movements as a clue of vitality because the attacker can easily simulate
IJCB2011
such a case, and also because real users often show a low vitality during
the authentication session.
In this paper, we perform both video and static analysis in order to employ
complementary information about motion, texture and liveness and
consequently to obtain a more robust classification.
Our approach Classification
AmILab's Spoof Detector implements a multi-clue approach. At classification stage scores are computed over a sliding window of a few
seconds of video.
Static analysis tackles the visual characteristics of a photo attack. Within this window, static analysis results in FxN scores (F frames and N
The visual representations that we propose to use are: Color and Edge visual representations). A unique score is computed through a DSC
Directivity Descriptor, Fuzzy Color and Texture Histogram, MPEG-7 algorithm. :
Descriptors (like Scalable Color and Edge Histogram), Gabor Texture, S sa = 1 min { S i , f }撃max { S i , f } i[1, N ] , f [1, F ]
Tamura Texture, RGB and HSV Histograms, and JPEG Histogram.
For each frame, each of the above mentioned visual representations result Finally, fusion between static and video analysis is performed as:
in a specific score.
Video analysis aims to detect vitality clues. Clues examined in this work
are motion analysis of the scene and the number of eye blinks that are
S =
{ 撃S sa1 S bl ,
1S sa 2S bl 3S m ,
if S m is high
if S m is low
represented by two independent scores.
S sa
Still Frame Characteristic analysis
D S
S
S bl C
Blink detection
Sm LOW?
Global motion Yes
Experimental results: the face spoof competition
For our experiments we used the Print-Attack Replay Database developed
for the IJCB 2011 Competition on counter measures to 2D facial spoofing
attacks from the Idiap Research Institute.
Although static analysis alone easily achieves a perfect separation in
the test set, we enhanced its classification with the video analysis in
order to grant performances even with higher quality printed photos or
high quality displays (smart-phones, tablets and other modern portable多
devices). 多
多
多
多
多
多
多
多
多
多
多
多
多
多= f k x i omega 多 ,
i
s多
ik
多=多
多= f k x i omega多 , S 多
k
i
s多
ik
多=多
S 多k
Introduction of video analysis results in lower performances in terms of
separation of scores' distributions. However, the proposed fusion scheme
still proved to be very effective and robust.
The contribution of video analysis in terms of robust classification will be
further investigated in future works.
Contacts
Ambient Intelligence Lab - Edificio 1, Loc. Piscinamanna, 09010 Pula (CA), Italy - Tel. +39 070 9243 2682
http://prag.diee.unica.it/amilab/ labiam@sardegnaricerche.it](https://image.slidesharecdn.com/amilabijcb2011poster-111031050151-phpapp02/85/Amilab-IJCB-2011-Poster-1-320.jpg)
Amilab IJCB 2011 Poster
- 1. Fusion of multiple clues for photo-attack detection in face recognition systems Roberto Tronci , Daniele Muntoni , Gianluca Fadda , Maurizio Pili , Nicola Sirena , Marco Ristori , 1,2 1,2 1 1 1 1 Gabriele Murgia , Fabio Roli 1 2 1 Ambient Intelligence Lab, Sardinia DistrICT, Sardegna Ricerche, ITALY 2 DIEE, Dept. Electric and Electronic Engineering, University of Cagliari, ITALY {roberto.tronci, muntoni, fadda, maurizio.pili, sirena, gabriele.murgia, ristori}@sardegnaricerche.it {roberto.tronci, daniele.muntoni, roli}diee.unica.it Introduction SARDEGNA RICERCHE We faced the problem of detecting 2-D face spoofing attacks performed by placing a printed photo of a real user in front of the camera. For this type of attack it is not possible to relay just on the face movements as a clue of vitality because the attacker can easily simulate IJCB2011 such a case, and also because real users often show a low vitality during the authentication session. In this paper, we perform both video and static analysis in order to employ complementary information about motion, texture and liveness and consequently to obtain a more robust classification. Our approach Classification AmILab's Spoof Detector implements a multi-clue approach. At classification stage scores are computed over a sliding window of a few seconds of video. Static analysis tackles the visual characteristics of a photo attack. Within this window, static analysis results in FxN scores (F frames and N The visual representations that we propose to use are: Color and Edge visual representations). A unique score is computed through a DSC Directivity Descriptor, Fuzzy Color and Texture Histogram, MPEG-7 algorithm. : Descriptors (like Scalable Color and Edge Histogram), Gabor Texture, S sa = 1 min { S i , f }撃max { S i , f } i[1, N ] , f [1, F ] Tamura Texture, RGB and HSV Histograms, and JPEG Histogram. For each frame, each of the above mentioned visual representations result Finally, fusion between static and video analysis is performed as: in a specific score. Video analysis aims to detect vitality clues. Clues examined in this work are motion analysis of the scene and the number of eye blinks that are S = { 撃S sa1 S bl , 1S sa 2S bl 3S m , if S m is high if S m is low represented by two independent scores. S sa Still Frame Characteristic analysis D S S S bl C Blink detection Sm LOW? Global motion Yes Experimental results: the face spoof competition For our experiments we used the Print-Attack Replay Database developed for the IJCB 2011 Competition on counter measures to 2D facial spoofing attacks from the Idiap Research Institute. Although static analysis alone easily achieves a perfect separation in the test set, we enhanced its classification with the video analysis in order to grant performances even with higher quality printed photos or high quality displays (smart-phones, tablets and other modern portable多 devices). 多 多 多 多 多 多 多 多 多 多 多 多 多 多= f k x i omega 多 , i s多 ik 多=多 多= f k x i omega多 , S 多 k i s多 ik 多=多 S 多k Introduction of video analysis results in lower performances in terms of separation of scores' distributions. However, the proposed fusion scheme still proved to be very effective and robust. The contribution of video analysis in terms of robust classification will be further investigated in future works. Contacts Ambient Intelligence Lab - Edificio 1, Loc. Piscinamanna, 09010 Pula (CA), Italy - Tel. +39 070 9243 2682 http://prag.diee.unica.it/amilab/ labiam@sardegnaricerche.it