You are on page 1of 4

A Survey on Efficient Approach for Determination for

Fake Biometric
Najuka Todekar

Prof. Pramod Mohod

Department of Computer Science and Engineering
GH Rasioni Women’s College of Engineering
Nagpur, India

Department of Computer Science and Engineering
GH Raisoni Women’s College of Engineering
Nagpur, India

Abstract—Biometric systems are vulnerable to spoofing attacks. A
reliable & efficient countermeasure is needed to prevent these
attacks. Liveness assessment is one of the techniques which is
very effective & can be applied to guard the system against
harmful spoofing attack. In this paper, a software based
protection approach is presented to prevent fraudulent access
attempts. This software based system computes various quality
measures extracted from one image depending on the specific
biometric modality to distinguish between legitimate and fake
samples. This makes the
system more efficient without
compromising the accuracy
Keywords-component; Liveness Detection, Image Quality
Assessment, Biometrics, Security, Spoofing, Fraudulent Attacks



Traditional methods for providing computer security such
as passwords, PIN etc; are based on the properties that can be
forgotten, disclosed, lost or stolen. e.g. Passwords often are
easily accessible to colleagues and even occasional visitors
and users tend to pass their tokens or share their passwords
with their colleagues to make their work easier. Thus these
methods do not authenticate the "user" as such. These methods
try to validate what someone remembers (e.g. Password/PIN
etc;) or what someone possess (e,g. ID card / token) instead of
validating the identity of the particular user. In last few years,
biometrics has been increasingly used in providing computer
security due to its distinct advantages over the conventional
authentication techniques. Biometrics refers to authentication
techniques that rely on measurable
physiological &
behavioural characteristics that can be automatically checked.
There are several types of schemes that can be deployed for
biometrics identification such as face, fingerprints, iris, hand
geometry, retina, gait etc; [2] Some of the key advantages
offered by biometrics over the traditional techniques are - a
scalable solution - reliable - increased accuracy - difficult to
breach However, with increased usage of biometrics, one of
the main threats faced by biometric systems of different
modalities is direct or spoofing attack. In the attack, the
attacker uses synthetic artifacts (e.g., printed images or
biometric modalities made of synthetic materials, gummy
fingers, face masks etc;) or tries to imitate the behaviour of
real user [10]. This helps the attacker to fraudulently access the
biometric system. So it is essential to propose and develop a
protection method against the threat. In this paper, different
approaches for identifying, whether the particular biometric

sample is real or fake, are discussed and an efficient approach
is described.


Researchers are focused on the design of specific countermeasures, that enable biometric systems to detect fake
samples, improving the robustness & security level of the
system. Besides other anti-spoofing approaches, special
attention has been paid by researchers & industry to the
Liveness Detection technique. It uses different physiological
properties such as odor, sweat etc; to distinguish between real
& fake trait. The key requirements that need to be satisfied by
liveness detection methods are : [26] (i) non-invasive - the
technique should in no case be harmful for the individual or
require an excessive contact with the user; (ii) user friendly people should not be reluctant to use it; (iii) fast - results have
to be produced in a very reduced interval as the user cannot be
asked to interact with the sensor for a long period of time; (iv)
low cost – a wide use cannot be expected if the cost is
excessively high; (v) performance - in addition to having a
good fake detection rate, the protection scheme should not
degrade the recognition performance (i.e., false rejection) of
the biometric system. Liveness detection methods are usually
classified into two groups Hardware based & Software based.
Hardware based techniques which add some specific device to
the sensor in order to detect particular properties of a living
trait such as blood pressure [3] , odor [4], heart beat [5]. These
hardware based techniques present higher prediction rate but
they are very costly and require complex implementation.
Unlike hardware based techniques, software based techniques
are generally less expensive & less intrusive as no extra device
such as spatial sensor is needed. Software techniques satisfy
almost all requirements of liveness detection mentioned above.
One of the approaches for discriminating fake fingers from
real ones [6] is based on analysis of skin distortion. The user is
required to move the finger while pressing it against the
scanner surface. The reference distortion code sequence is
obtained from a well-trained user with a uniform & smooth
movement. It was able to correctly represent most of real
finger distortions. This approach is privacy-friendly & does not
require additional expensive hardware besides a finger print
scanner capable of capturing & delivering frames at proper
rate. It achieves better results when the users are well- trained
and habituated. However this approach requires additional
training to the users as well as multiple image frames in order
to get better results. In another approach [7], which is based on
quality measures, the liveness detection method presented has

Similar to finger. and it is not rare that fingerprint images captured from a gummy finger present local gaining artifacts such as spots and patches. This bio-metric method is applicable to multi-modal modalities. Moreover. spoofing attack with Finger.. found in both type of images. However. The two types of methods have certain advantages and disadvantages over the other and. but it is not useful in fake fingerprint detection.based methods can protect the system against the addition of reconstructed or synthetic samples onto the communication channel between the sensor and the feature extractor [11]. It reduces the computational load. For added advantage that it requires just one image from a finger to decide whether it is real or fake. on general.) are done prior to calculation of the IQ features. measures like the mean squared error respond additional to additive noise. general artifacts. face images captured from a mobile device will most likely be over. In the same way. iris detection or face extraction etc. from the image and then decides whether the image presented is real or fake. In this work. All these help to solve the problem of spoofing detection. Liveness detection methods represent a difficult engineering problem as they have to satisfy certain challenging requirements (i) user friendly. and not from the characteristic itself). Different quality measures present diverse sensitivity to image artifacts and distortions. results have to be generate on a very less time interval as the user cannot be asked to interact with the sensor for a long period of time. in an ultimate attack in which an unnaturally produced image is directly injected to the communication channel before the feature extractor. software. an input biometric sample is classified as either real or fake. Also. As a common comparison. For example. the amount of occlusion of the eye is applicable in iris anti-spoofing mechanism. software techniques may be embedded on the feature extractor module which makes them potentially accomplished of detecting other types of illegal break-on attempts not necessarily classified as spoofing attack. It has no pre-processing steps (e. For example. Liveness detection using image quality assessment is based on hypothesis that : “It is expected that the quality of the fake image taken in an attack will be different from a real image”. hardware-based schemes generally present a higher fake detection rate. iris images taken from a printed paper may be more unclear or out of focus. Complementarity 3. METHODOLOGY Liveness detection methods are generally classified onto two types (see Fig. Complexity 4. In fake biometric detection system. a combination of both would be the most advantageous protection approach to increase the security of biometric systems. amount of information found in images.or under-exposed.”[1] Predictable quality differences between real and fake samples may contain: color and luminance levels. A general diagram of fake detection system is shown in Fig. In the proposed system. Image manipulation detection [9] has been successfully carried out in previous work. a large use cannot be expected if the cost is very high. This fact shortens the acquisition process & reduces the inconvenience for the final user also eliminates the need for training the users. there is no biometric system in general which uses image quality as a protection method e. The features used. The main role of the method is to find out a set of features which help to construct a suitable classifier. local artifacts. this fake sample will most likely lack some of the properties found in natural images. For example. pixel intensity etc. gummy finger may contain spots or patches. as they run directly on the acquired sample. the sample is classified as real or fake using classifiers like Linear discriminant analysis (LDA) & Quadratic discriminant analysis (QDA). Performance 2. PROPOSED WORK The use of image quality assessment for liveness detection is motivated by the supposition that: “It is expected that a fake image captured in an attack attempt will have different quality than a real sample acquired in the normal operation scenario for which the sensor was designed. luminance levels. It uses both Full Reference & No Reference image quality measures which are classified according to 4 general criteria: 1. they fail to generalize as they are designed to work on specific modality & usually used to detect one specific type of spoofing attack. degree of sharpness. IV. iris images captured from a printed paper are more likely to be unclear or out of focus due to trembling. structural distortions or natural appearance. These fake samples lack some of the properties in the original image. liveness detection scheme for Iris based on quality related measures has been presented. and less intrusive since their implementation is clear to the user. Iris or face images is considered as a type of image manipulation that can be successfully detected to a certain extent.g. (iii) low cost. which add some particular device to the sensor on order to detect Exacting properties of a living feature. to detect certain fingerprint spoofs. Liveness detection techniques use different physiological properties to differentiate between real and fake sample. power spectrum. while others such as the spectral . III.g.e. quantity of information. do not calculate any specific property of a given biometric modality or of a specific attack. and quantity of sharpness. on this type the fake trait is Detected once the sample has been acquired with a normal sensor (i. features used to differentiate between real and fake traits are extracted from the biometric sample. 1): (I) Software-based techniques. (ii) fast. fingerprint segmentation. This method considers the whole image rather than considering any trait specific properties. In this approach [8]. General image quality assessment is used as a protection method against different biometric attacks. Speed General image quality measures are fast to compute and easy to combine with simple classifiers. at the same time software-based techniques are on general less expensive (like no extra device is needed). an Iris image can also be used for liveness detection.The potential of general image quality assessment as a protection method against different biometric attacks (with special attention to spoofing). people should be averse to use it. It extracts the necessary features such as local angle. 2 The system uses only one input as a biometric sample in order to keep its generality & easiness. face images taken from mobile may be over-exposed or underexposed. [1] Real & fake samples have quality differences in color. Once the feature vector has been generated. structural distortions. a novel parameterization using 25 general image quality measures are considered. ridge & valley frequency may be used as a good parameter for measuring but it cannot be used in iris liveness detection. it calculates on any image. (II) Hardware-based techniques.

271–276. S. (2012). Ortiz-Lopez. Bayram. 1992. a novel. Dec. F. Prabhakar. pp. a) Assessment of video quality measures.” EURASIP J. ” IEEE Trans. Wang. R. “Image Quality Assessment for Fake Biometric Detection: Applicat ion to Iris. D. Image Process. comparative results with other existing solutions d. Hence in the proposed system. 041102-1–041102-17. softwarebased protection measure is proposed to protect the system against fraudulent biometric system access attempts. Apr. Signal Process. 13. E. and A. Proposed scheme Based on the existing research. and E. Lapsley. 2003 P. W. 1. Jain. vol. M. veins. Future enhancement for this work can be. 33–42. fraudulent attacks will be different. 1995.utexas. 2. “ECG analysis: a new approach in human identification”. Therefore. 2008. A. B. Sankur. Electron.. 4.” Electron. “A high performance fingerprint liveness detection method based on quality related features. I. 44.” Signal Process. 808–812. Lu. Alonso-Fernandez.ece. no. The flowchart below illustrates the high level functioning of the system. Avcibas. “Fake fingerprint detection by odor analysis”. Comput. Lett. using a large range of IQMs exploiting complementary image quality properties should allow detecting the aforementioned quality differences between real and fake samples expected to be found in many attack attempts (i. validation of new biometric system b. Pankanti. IEEE ICIP. Dario Maio . A. IEEE Transactions on Information Forensics and Security. the performance of the system can be optimized without compromising accuracy. Philipson L. 1998 Denis Baldisserra .difference” theory and that image quality measures have the possible to achieve success in biometric protection tasks. Fierrez. utilize publically available databases for reproducible evaluation of multiple biometric traits The proposed research work presents software-based fake biometric detection method by identifying optimum set of image quality measures and thus provides an efficient approach for evaluation of multimodalities.. This software-based method computes relevant features based on the specific modality from more than 25 image quality features of biometric sample to verify its legitimacy. Ghanbari. T. VOL. Eskicioglu and P.” IEEE Trans. Nill and B. and Julian Fierrez. and A. no. “Scope of validity of PSNR in image/video quality assessment. demonstrate “Image quality assessment” as a technique for securing biometric systems against various kinds of attacks c. Fisher. vol. D. 23. and M. and J. 600–612. Ortega-Garcia. 12. 8. 50. 2012. B. vol. B. S. Image Commun. Memon. 4. Annalisa Franco . pp. Some of the key contributions are a. 2012. pp. Imag.. Lin. “Objective image quality measure derived from digital image power spectra. 2012 S. 2..” IEEE Trans. 2002. Yao. “Iris liveness detection based on quality related features. “Image quality measures and their performance. Liveness assessment can be applied to the biometric systems to protect them against harmful attacks.. 2. S.. Ong. 15. ” in Proc.htm . In the current research.. NO. IEEE TRANSACTIONS ON IMAGE PROCESSING. pp./Apr.e. J.. K. J. pp. J. “Fake finger detection by skin distortion analysis”. 2005. J. vol. there will be a provision to configure the system for a specific modality as per the specific implementation. no. Adv. [1] REFERENCES [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] Figure 1. 2006. [12] [13] [14] [15] [16] CONCLUSION AND FUTURE WORK In the last few years. vol. Liu. 813– 825. N. V. N. while gradient-related features respond to distortions concentrated around edges and textures. Cappelli . and J. 397–400. 21. Sayood. 875–882. Avcibas. and N. and Z. pp. Villarini. A. 2008. no. Bovik. and B. 2959–2965. C. vol. 5th IAPR ICB. “Image quality assessment based on gradient similarity. Davide Maltoni. 2008. vol. and K. Wide P. It is observed that image quality properties of real access & [17] [18] [19] [20] Javier Galbally. 4. 13. currently identified 25 parameters may be extended further by considering following additional quality measures [27] [28] [29] [30]. This would result in optimum performance of the system without impacting the accuracy. 2006 A. Hewage. Pettersson O. no. Galbally. US Patent. vol. 11. “Contrast signal-to-noise ratio for image quality assessment. In this way.. Hoffman. no. pp.” IEEE Security Privacy. Apr.. 363-373. 27. Fierrez. pp. Nandakumar. 2001. no. D. b) Further added modalities such as hand geometry. H. pp. Commun. “Image quality assessment based on edge preservation. “Statistical evaluation of image quality measures.” J. K. Springer LNCS-3832. 800–801. FEBRUARY 2014 S.phase error are extra sensitive to blur. pp. “Biometric template security.1. pp. Biel L. 1500–1511. Jan. Lin. Pare. I. In this paper. K. Image Process. Eng. given that the technique with multi-attack protection capabilities). “Image manipulation detection. 265–272. pp. Lee. “Image quality assessment: From error visibility to structural similarity. Syst. 1. W. no. M. 206–223. Mar. Sankur. it is determined that certain quality measures are more relevant for certain modalities. Available: http://live. Bouzas. vol. 2006. Narwaria. 28. Jain. the analysis of risks of various types of attacks against biometric systems has been a prominent field of research. R. Ortega-Garcia. Galbally. G. vol. vol. Electron. 2012 J.” in Proc. Q. A. c) Extension of new image quality measures implemented Existing work has made contributions to the state of the art in the field of biometric security. pp. J. Z. Antonelli . no./Apr. Fingerprint and Face Recognition”. Martini. Imag.” Opt. Simoncelli. pp. Huynh-Thu and M. “Biometric recognition: Security and privacy concerns.. Mar. vol. “Anti-fraud biometric sensor that accurately detects blood flow”. Nagar. 113–129. Sep. This leads to a great progress in the area of safety for biometric based applications to protect biometric systems from vulnerabilities like spoofing attacks. P. LIVE [Online]. on Instrumentation and Measurement. no. Sebastien Marcel. vol. Sheikh.” in Proc. IAPR ICB.” IEEE Trans. 2004.. 31. Maltoni. pp. Maio .” Future Generat. C. So consider that there is sound proof for the “quality.” J. 4.

2009 [25] A. 17. 2002 [24] X. [27] D. Image Process. Mar. [22] R. 111–119.Prabhakar. Zhu and P. Farid. C. “No-reference perceptual quality assessment of JPEG compressed images. Mar. “Image information and visual quality. Sheikh. Maio. pp. IEEE ICIP. Wang.” IEEE Signal Process. “RRED indices: Reduced reference entropic differencing for image quality assessment. Multimedia Exper. 21. Mittal. pp. “Steganalysis using higher-order image statistics. Bovik. 477–480. vol. USA: Springer-Verlag. 15. pp. no. 513–516. C. pp.. 2006. “A no-reference sharpness metric sensitive to blur and noise. Feb. pp. “Handbook of Fingerprint Recognition”. A. 20.” IEEE Trans.[21] H. D. Bovik. 2013. Feb.” in Proc. “Making a „completely blind‟ image quality analyzer. . Sep. C. Lett. vol. C. Workshop Qual. R. “A two-step framework for constructing blind image quality indices.” IEEE Trans. [26] A. 2. Maltoni. 2. H. no. C. no. 2006. Forensics Security.. no..” in Proc.. 5. Moorthy and A. R. NY. 3. 64–69. K. and A. 517–526. Soundararajan and A. 1. 2012. Sheikh and A. Bovik. Lett. pp.Jain and S. Milanfar. Image Process. vol. Soundararajan. [23] Z. 430–444. Bovik. Inf. pp. May 2010. New York. Int. vol. vol.” IEEE Signal Process. and A.. 2009 [28] S. Lyu and H. Bovik. 209–212. 1.” IEEE Trans. no. R.