CLIAMS:We Claim:

1. A method for detecting liveness in fingerprint images which have passed genuineness test by using a minutiae based fingerprint marcher, the method comprising steps of:
subjecting the fingerprint images to steerable wavelet packet based fingerprint liveness detection for extracting features from textured patterns;
dividing the fingerprint images into local blocks of same size;
applying Fourier bases on local blocks of the fingerprint images for generating Fourier expansion within each of the block;
applying Fourier framework for obtaining two dimensioned wavelet bases;
filtering the fingerprint images for removing conjugated parts;
refining the fingerprint images by using steerable wavelet packets for analyzing the fingerprint images using directional zoom-in features for liveness detection;
applying inverse Fourier transform on local Fourier basis for obtaining biorthogonal bases;
performing directional analysis on each of the fingerprint images; and
performing second expansion using finer grind on the fingerprint images for determining liveness.

2. The method as claimed in claim 1, wherein the Fourier framework enables for creating two dimensional wavelet bases using tensor product of one dimensional bases.

3. The method as claimed in claim 1, wherein the filtering of the fingerprint images for removing conjugated parts is done by two wavelets, which form an approximate Hilbert pair.

4. The method as claimed in claim 1, wherein the second expansion is done by using steerable packet expansion for calculating and matching the finer grid at finer resolution for determining liveness in the fingerprint images.
,TagSPECI:Field of the invention

The present invention relates to a method for detecting liviness in fingerprint, more particularly, the present invention relates to novel steerable wavelet packet based method for fingerprint liveness detection.

Background of the invention

Among all biometric identifiers, fingerprints are widely accepted and used. Once compromised, it’s compromised forever. Thus biometric information is extremely sensitive and is required to be protected while operating in real situations. It is thus very much desired that bio- metric systems are secured and matured before they are deployed in real-life situations.

Spoofing

Fraudulent entry of any unauthorized person into a finger- print recognition system by using faux fingerprint sample is termed spoofing. These unauthorized entries may provide access to private information of a person or company causing serious damages. In the recent past, various spoofing techniques have been reported, including fake fingers using gelatin (gummy fingers), moldable plastic, clay, play-doh, wax, and silicon, developed from casts of live fingers or latent fingerprints. Cadaver fingers have also been shown to reliably be scanned and verified by fingerprint devices of various technologies.

Many countermeasures have been proposed against such attacks, while some are hardware based some other are software based with an aim of ensuring that the information provided to the scanner is not only genuine but is also coming from live source.

Various spoofing exercise was conducted on three different Unique Identification Authority of India (UIDAI) approved devices. The spoofing exercise was conducted using cheaply available materials like play-doh and white silica for preparing fake fingerprints. The quality was assured by pruning all the spoof images with NFIQ (NIST Fingerprint Image Quality, where as NIST stands for National Institute of Standards and Technology) score of 5. The spoofing results are shown as bar charts in Figure 1. In these, the enrollment was assumed to be using live genuine samples and system was tested using spoof image samples. The data distribution is as shown in Table (I) and all the three scanners were tested using the spoofs. Even though the procedures were not optimized and the fake fingerprints were not created with great care and precision, spoofing rates of 92%, 97% and 95% spoofing accuracy for CrossMatch, Secugen and L1ID make UID respectively was observed. This demonstrates need to have ‘Liveness’ tests in place for better security of system and better privacy preservation for the users of such devices.

Liveness Detection Techniques of Prior Art

Previously, liveness detection has been suggested as a countermeasure against spoofing of fingerprint scanners. Several liveness measures including pulse, pulse oxime- try, electrocardiogram, temperature detection, and perspiration based spectroscopy based techniques are suggested.
Liveness detection for fingerprint scanners can be applied using:-

• Extra hardware to acquire liveness signs; and

• Soft processing techniques to extract liveness signs from already captured information.

Few of the most accepted liveness methods are as follows:

1. Pulse oximetry method is based on differential absorption of two wavelengths of light projected through the finger. While blood oxygen content is ignored, the pulse information is used for liveness detection. This method suffers with drawbacks including a requirement that the finger be completely covered for the test since ambient light may interfere.

2. Another suggested method is an electrocardiogram (ECG) measurement. This method requires two contact points on opposite sides of the subject’s body to be able to perform the measurement. This method is bulky, and hence less practical.

3. An elegant method uses multi-spectral sensors to ex- pose the finger to different optical wavelengths. This method requires extra hardware in a specially designed fingerprint scanner. A gelatin faux fingerprint is shown to have optical properties very similar to human skin, which this method shorts fall of make distinction.

4. Another method detects temperature of epidermis which typically is in the range 25 - 30o C. Main drawback of this method is vulnerability increases as the range of operating temperatures increases.

5. A US patent 5737439 - ‘Anti-Fraud Biometric Sensor that Accurately Detects Blood Flow’ by Smart Touch LLC describes a method to determine finger blood-flow. The technique is based on pulse-oximetry.

6. Yet another technique based on pulse oximetry is mentioned in “S. Kurt, “Biometrics - what you need to know,” Security Portal 10, Jan 2001”, but the technique can be deceived by using a translucent artificial fingerprint as mentioned in by S. Schuckers, “Spoofing and anti-spoofing measures,” in Information Security Technical Report, vol. 7, 2002, pp. 56–62”.

7. The electrical resistance of human skin with pre- specified range is also used as liveness measure in liveness. This method can be spoofed by gelatin fingers as they have comparable moisture level as compared to human skin.

8. Further, perspiration pattern based method makes use of pattern that gets evolved over time and relies on multiple images captured.

The liveness algorithms mentioned above are either of being costly, bulky, hardware dependent and environment dependent. To overcome these issues and help easy deployment of the ‘liveness’ tests into real practices there is a need to provide a new method of liveness detection for fingerprints.

Objects of the invention

Object of the present invention is to provide a method for detecting fingerprint liveness, which prevents fake entries by providing more accurate match results as compared to the existing method.

Another object of the present invention is to provide a method for detecting fingerprint liveness, which uses hardware from the existing devices and does not require any additional hardware.

Summary of the invention

According to the present invention there is provided a method for detecting liveness in fingerprint images which have passed genuineness test by using a minutiae based fingerprint marcher. The method having steps of subjecting the fingerprint images to steerable wavelet packet based fingerprint liveness detection for extracting features from textured patterns. Thereafter, the fingerprint images are dividing into local blocks of same size. After that, the Fourier bases is applied on local blocks of the fingerprint images to generate Fourier expansion within each of the block. Thereafter, Fourier framework is applied to get two dimensioned wavelet bases. Further, the fingerprint images are filtered to remove conjugated parts. Then, the fingerprint images are refined by using steerable wavelet packets for analyzing the fingerprint images using directional zoom-in features for liveness detection. Thereafter, inverse Fourier transform is applied on local Fourier basis to obtain biorthogonal bases. Then, directional analysis is performed on each of the fingerprint images. At last, second expansion is reformed using finer grind to determine liveness.

Brief description of the invention

The advantages and features of the present invention will become better understood with reference to the following detailed description and claims taken in conjunction with the accompanying drawings, wherein like elements are identified with like symbols, and in which:

Figure 1 shows a bar chart of spoofing results of existing figure print scanner;

Figure 2 shows a flow chart of a method for detecting liveness in a fingerprint image in accordance with the present invention;

Figure 3 shows a basis functions for four quadrants and four directions of the fingerprint image;

Figure 4 shows a selective orientation analysis performed by steerable wavelet packets;

Figure 5 shows a second expansion with finer grid with four lattice squares around an origin characterize a DC part and other squares correspond to higher frequency textures;

Figure 6 shows an energy maps of steerable wavelet packet representations of live and spoof impressions of 50 subjects; and

Figure 7 shows a Receiver Operating Characteristics (ROC) curves for steerable wavelet packet based fingerprint liveness system, wherein AA, BB and CC are ROC curves when the system was tested without the method for detecting liveness and A, B and C with the method for detecting liveness for CrossMatch (CM), L1ID and Secugen respectively.

Details description of the invention

For a thorough understanding of the present invention, reference is to be made to the following detailed description, including the appended claims, in connection with the above-described drawings. Although the present invention is described in connection with exemplary embodiments, the present invention is not intended to be limited to the specific forms set forth herein. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but these are intended to cover the application or implementation without departing from the spirit or scope of the claims of the present invention. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting.

The term “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced item.

The present invention provides a method for detecting fingerprint liveness. The method prevents fake entries by providing more accurate match results as compared to the existing method. Further, the method uses hardware from the existing devices and does not require any additional hardware.

Referring now to figure 2, a flow chart of a method (100) for detecting fingerprint liveness in according to the present invention is illustrated. The method (100) includes steps of:

At step (10) the method (100) starts.

At step (12), fingers that need to be identified are scanned to collect fingerprint by using various UID approved devices. For the purpose of explanation only UID approved devices, such as Secugen, Cross Match, L1ID and the like are be used for collecting fingerprints. It may be obvious to a person skilled in the art to use other similar devices of different manufacturers.

At step (14), the fingerprints are passed though NIST fingerprint quality engine (NIFQ) score test. The fingerprints having NIFQ score 5 are filtered and the fingerprint having score less than 5 are selected to create a database. The data distribution is shown in the Table I, where total intra and inter class combinations are clearly stated. C M stands for Cross Match and SH (I I ) stands for Secugen Hamster (II) scanner.

Table I
Data set: Distribution

Sensor Live
Classes Spoof
Classes Sessions
per class Images per
Class Intra class
per class Inter class
per class
L1ID 52 52 5 10 63,700 6630000
CM 52 52 5 10 63,700 6630000
SH(II) 52 52 5 10 63,700 6630000

At step (16), the database is subjected to verified query. Here verified query refers to an attempt made by a user while trying to pass verification test of the device for which he or she is already enrolled.

At step (18), the fingerprints in the database are subjected to a minutiae based fingerprint marcher. The minutiae based fingerprint marcher is a stander device available as on date and use of which is obvious to a person skilled in the art. This marcher provides identifies each of the image as imposter or a genuine match. The genuine matches are referred as perfect match. This minutiae based fingerprint marcher is used to segregate the image at the first level. The genuine images are further used for liveness test.

The improvement of the method (100) of the present invention starts at step (20). At step (20), the fingerprint images obtained from the minutiae based fingerprint marcher are subject to steerable wavelet packet based fingerprint liveness detection, in which features such as directional wavelet coefficients and texture patterns of each of the images is analyzed. As the fingerprint images obtained from the minutiae based fingerprint marcher are rich in oriented textured ridge patterns, therefore edges and textures in the fingerprint images can exist at vivid possible locations, scales and orientations. The ability to efficiently analyze and extract features from textured patterns is thus of fundamental importance for building liveness embedded robust fingerprint recognition systems.

The simplest model of patch of periodic and cyclic texture located at (x0 , y0) ( x0 , y0 represents translation parameters) is provided by a windowed complex exponential,

w(x - x0 , y - y0 )ei(?x+?y) ----- (Equation - 1)

where “w” is a functional localization around the origin. Locating patch of periodic and cyclic texture as the fingerprint images are highly oriented and cyclic.

At step (22), each of the fingerprint images is divided into local blocks of same fixed size.

Thereafter, at step (24), Fourier bases (z) is applied to each of the local blocks of the fingerprint images to encode the entire image by virtue of generating Fourier expansion within each of the block of the fingerprint images. Local Fourier basis is used for obtaining the most appropriate representation for texture analysis. The basic problems in this easiest approach are as follows:
• The size of the block should be adapted to the fingerprint ridge map content. (a large geometric feature should NOT belong to several small blocks etc.)
• The size of the blocks should be adapted to the frequencies of complex exponentials. (shorter blocks for higher frequencies etc.)
• ‘Blocking’ artifacts at boundaries of blocks
• Difficulty to superimpose blocks of different sizes.

At step (26), to solve above referred issues Fourier framework is replaced with multire solution framework. Two dimensional wavelet bases are created using tensor product having one dimensional bases. Let f be the scaling function and ? be the corresponding wavelet function, four wavelet functions can be written as:

f(x)f(x), if k=0
?k (x, y) = f(x)?(x), if k=1 ----- (Equation - 2)
?(x)f(x), if k=2
?(x)?(x), if k=3

At step (28), the fingerprint images are filtered to remove conjugated parts. The associated filter banks mk (?, ?), k = 1, 2, 3 can resolve 2.5 directions namely, horizontal, vertical and an undecided diagonal direction. A wavelet packet strategy is deployed to adaptively construct an optimum tiling of the plane. The implementation of the wavelet packet strategy as represented in Equation 3, and its geometric interpretation, however, becomes very challenging as tensor product of two real valued wavelet packets gets associated with four symmetric peaks in the frequency plane. The main challenge lies with the fact that intensity in the image is either oscillating as a planar wave ei(wx x+wy y) or with the conjugate frequency ei(wx x-wy y) .

m3 (?, ?) = 0 if ? > 0 and ? < 0, or if ? < 0 and ? > 0 -- -- -- (Equation 3).

In order to construct such filters two wavelets ?gand ?h which form an approximate Hilbert pair are used:
--- -- -- (Equation 4)

with fh , fg being the corresponding scaling functions.

To avoid localization of Fourier transform to only one quadrant for tensor product like ?h (x)?g (y), tensor product for wavelet ?h is taken as follows:

?h,1 (x, y) = fh (x)?h (y)
?h,2 (x, y) = ?h (x)fh (y) --- -- -- (Equation 5)
?h,3 (x, y) = ?h (x)?h (y)

Similar tensor products are used for ?g. Sum and difference were calculated:

?i (x, y) = ?h,i (x, y) + ?g,i (x, y)
?i+3 (x, y) = ?h,i (x, y) - ?g,i (x, y), i = 0, 1, 2 -- (Equation 6)

This transform gets resolved into six different directions with four directions of greater significance.

At step (30), the transformed images are further refined using steerable wavelet packets. The steerable wavelet packets are used for analyzing the fingerprint images using directional zoom-in features for liveness detection. Basis function for four quadrants and four directions is shown in Figure 3.

The steps 22, 24, 26, 28 and 30 together provide steerable wavelet packets for analyzing the fingerprint images using directional zoom-in features for liveness detection in totality.

The step 28 and 30 enables checking of pores for the liveness by perspiration activity/pattern in the fingerprint image.

At step (32) biorthogonal bases is obtained by using unitary nature of Fourier transform by applying inverse Fourier transform on local Fourier basis.

Let f ? L2 (R) and let fˆ be the Fourier transform of f . Cover of the frequency
axis will be, un,k and u˜n,k .

-- --- equation (7)
-- -- equation (8)
In equation (8), ?n,k (x) is a complex valued function and has a phase component which captures the orientations, “i” is the translation index and ln is the analysis scaling factor. This theme is extended to a two dimensional kernel by partitioning the frequency plane through lattice squares.
-equation 9
(?m , ?n ) is the center of each rectangle of size hm × ln .

At step (34) directional analysis is performed on each of the fingerprint images. ?m,j ? ?n,k for different values of frequency fixed hm and ln is used as basis function for analysis. The fingerprint images used for analysis were of size 320 ×320 and the analysis windows were defined by hm = ln = 16, d = 8.

Figure 4 illustrates the selective orientation analysis performed by the steerable wavelet packets. As shown in figure 4, at a given scale this method can resolve many more orientations than standard wavelet packet. As scale gets coarser (smaller hm; jn), it is possible for resolving more directions. These representations come out of four sets bases with orientations

At step (36), a second expansion is performed using finer grid. Each quadrant was further divided into four quadrants. The second expansion is done by using steerable packet expansion for calculating and matching the finer grid at finer resolution. This is shown in Figure 5. This finer grid is used further for characterizing the liveness score. The L2 domain again helps us carry out the energy analysis of the fingerprint images. The liveness score is calculated using following formulae:

Liveness Score = hm ln e2ip(?m x+?n y) -- -- -- (equation 10)

The images are defined live only if the images successively pass abovementioned authentication routines.

The method ends at step (38).

In order for deciding the threshold on Liveness Score for deciding if the sample fingerprint is live or not, an experiment was conducted on 50 subjects with their live and spoof images. The energy maps of these 50 subjects were drawn for deciding threshold value. A dynamic way of deciding the threshold remains future work. The energy maps are shown in Figure 6.

The data distribution is as shown in Table (I). For all the three scanners tests are carried out by invoking devices using spoof as well as live data, once in absence of the method (100) and then by incorporating the method (100). Correspondingly six ROC (Receiver Operating Characteristics) curves are shown in Figure 7. Receiver Operating Characteristics (ROC) curves for steerable wavelet packet based fingerprint liveness system. AA, BB and CC are ROC curves when the devices were tested without using the method (100) for CrossMatch (CM), L1ID and Secugen respectively is shown in figure 7. AA, BB and CC curves have EERs of 35:08%, 39:85% and 45:37% respectively. A, B and C are ROC curves when the devices were tested with the method (100) for CrossMatch (CM), L1ID and Secugen respectively. A, B and C curves have EERs of 0:09%, 1:18% and 2:65% respectively.

The method (100) is treated as two class problem. The ‘genuine’ class indicates both passing the recognition test as well as liveness test. Another class shall have genuine + spoof, imposter + live and imposter + spoof combinations and together will be called as an ‘imposter’ class as shown in figure 2.

For CrossMatch, from AA (EER=35.08%) to A (EER=0.09%) indicate substantial improvement. EER of AA suggests existing scanners vulnerable for spoofing attacks and EER of A indicate effectiveness of the method (100). Similar results are observed for other two scanners as well.

In this disclosure, a steerable wavelet packet based fingerprint liveness detection the method (100) is designed, implemented and tested across three different fingerprint scanners. Initially white silica based spoofing method is introduced. Spoofing was shown with more than 90% accuracy across all three scanners. The tests were carried out both with and without liveness and average improvement of 36.33% in EER was observed. The developed matching scheme is tested for the high resolution data (686 ppi) for 114 live and spoof fingerprint classes. ROC is plotted and EER of 2.97% is obtained.

The method (100) of the present invention has advantage of preventing fake entries and provide more accurate results as compared with the existing method for detecting fingerprint liveness. Further, the method (100) uses hardware of the existing devices, therefore does not require alteration or addition of any component.

The foregoing descriptions of specific embodiments of the present invention have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the present invention to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the present invention and its practical application, and to thereby enable others skilled in the art to best utilize the present invention and various embodiments with various modifications as are suited to the particular use contemplated. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but such omissions and substitutions are intended to cover the application or implementation without departing from the spirit or scope of the claims of the present invention.