The identification method of biometric images is developed where the filtration using Ateb-Gabor was implemented. Attacks that act on a biometric system were analyzed, and an attempt to counter these attacks was made. While applying skeletonization, we proposed to use a wave algorithm. It forms a ridge of minucius in the center of mass. This method of filtering has more opportunities than the traditional Gabor filter. A new filtering method has been developed that extends the existing filtering methods for prints. The method is based on Ateb-filtering, which increases the capabilities of standard filters because it is based on differential equations with degree of nonlinearity. Since it is based on Ateb-functions, which have more extensive properties than classical trigonometry. Such kind of filtration gives a solution to a problem that is clearly defined in specific parts of the area, both in the spatial and in frequency domains.
The most common static method of biometric identification is the comparison of
fingerprint [
Biometric technologies are very vulnerable to hacking attacks. Because hackers can
crack the biometric passport chip and access the information stored on it. Source [
Specifically, there were attempts to break and clone a biometric passport of a US
citizen. You can record any arbitrary information on the chip or block it completely
[
Another perspective is medical. Hackers can hack medical devices implanted in a
person's body. By breaking, for example, the Merlin @ home system [
Thus, today, the effectiveness of biometric technologies in the context of foreign
policy security seems controversial. The development of modern information
technologies makes it possible to bypass the security system, which puts new tasks before
information security [
Human control and surveillance [
Interference with the privacy of citizens [
The problem of storing biometric databases of citizens [
Impact of biometric technologies on human health [
Hackers continually have to invent new ways to deceive biometric scanners in order to log on. Therefore, we have synthesized different types of attacks to use them as a means of hackers' counteracting.
Biometrics-Based personal authentication systems, particularly fingerprints,
become more popular than traditional systems that are based on tokens (keys or
password) [
Traditional authentication systems are not prepared to distinguish between impostors who have illegally acquired the privileges to access a system and the genuine user. Furthermore, biometric systems can be more user-friendly because there is no need for the user to remember passwords.
Regardless of these benefits, biometric systems have some disadvantages. That is to say that biometric systems are vulnerable to external attacks, which could reduce their level of security.
In Ratha [
The first type of attack involves submitting a fake biometric fingerprint sample for scanning. In other words, the hacker submits pre-intercepted biometric data. The second type is known as “replay attack.” In some way, hackers reproduce a biometric sample and enter the system. In the third type of attack, the recognition module provides false values of the feature which were chosen by the intruder. In the fourth type of attack, the values of a specific function are replaced by those who were selected by the hacker. The set of recognition features can be modified to obtain an artificially high matching score in the fifth type of attack. The attack on the database by adding new templates, modifying existing templates and removing existing templates carry out in attack number 6. The attack in the seventh type carries out when a template is broadcasting through a communication channel between the system database and the matcher module, resulting in a change of templates in the database. Finally, an intruder might redefine the result of the matcher (accept or discard).
The algorithm of segmentation and enhancement of the fingerprint image, which uses
a new method of filtering based on Ateb-functions [
Step 1. Image normalization. At this stage, we carry out the scaling process of imprint to uniform scale and geometric sizes with a clearly defined resolution.
Step 2. Computation of the local orientation. It means the scaling process of imprint to the origin of coordinate and turning the imprint with the setting of the origin of the coordinates and the polar axis. While scanning, we incline fingers at any angle. It is necessary to have an apparent reference to the coordinate system in order to recognize the imprint. We try to find out it in the second step.
frequency matrix based on the normalized and orientation image, which was performed in steps 1 and 2.
Step 4. Imprint segmentation. The construction of an imprint mask by breaking
down the normalized image into blocks and performing the classification task of each
block, dividing them into those who contain and not contain backbones. After this, we
smooth the mask by Gabor filtering [
Filtration of the normalized image. The outlines of parallel ridges and valleys with
well-defined frequency and focus on the fingerprint image contain the information
that helps eliminate objectionable noise. A bandpass filter [
There are three parameters for using Gabor filters to the image: the frequency of the Ateb-function f wave, the filter direction, the mean square deviations of the Gaussian shell x 'and y'.
Step 5. Filtration of the normalized image. We apply a set of Gabor filters,
AtebGabor [
We calculate three values for each pair of such points: the module of the vector, which connects a pair of minucius, the vector orientation relative to the horizontal, and the directions of the papillary lines with minucius relatively to the horizontal. So, the template contains a description of the imprint in relative units, which neglects alters the image orientation.
The algorithm evaluates the percentage of matches between the corresponding three values. The speed of response of systems is determined by the execution of several significant operations - exponentiation, computation of a root, division, calculation of arctangent.
4
The next stage involves making the fingers of the papules thinner and bringing them to a thickness of one pixel of curves. In this case, the contour is selected, which is highlighted in Fig. 5 in a darker color and the construction of a graphic representation of curves. We will call the contour of the image a collection of its pixels, around which there is a jump-like change in the brightness function. The image contours will be represented by lines in one pixel wide. If, in addition to areas with constant brightness, there are areas with brightness that is smoothly changed in the original image, then there is no guarantee of continuity of the contour lines when specifying the contour lines: the discontinuities in those places where the change of brightness function is not sharp enough will be observed.
On the other hand, noise is present on a piecewise permanent image, and then unnecessary contours may be recognized that are not desirable when creating the boundaries of the domains. The algorithms for selection of contours are developed, and the behavior of contour lines is taken into account. Unique extra algorithms can eliminate gaps and eliminate excess contour lines.
For the selection of boundaries, i.e., brightness variations are made by the wave
method [
Then, the image is split into individual fixed blocks. On the curve we find the points of maximum or minimum. A skeleton image is created.
By the skeleton of the image, we evaluate the characteristic points that were given in Fig. 2 and 3. 5
Recognition and identification a d g j m b e h k n c f i l o
It is necessary to highlight the critical points after the formation of the skeleton of a
biometric image and compare them to the key points which exist in the database in
order to carry out the identification in this way. The main characteristic points
highlight the graphic image. We suggest installing them in the range from 12 to 24. In the
event of a large number of points, there probably will not be enough computational
resources, but when there are few points, there is the probability of admitting
someone else's fingerprint. Owing to this, the algorithm has two parameters: FAR - the
access error of extraneous user, which should be 0,01%, and FRR - a mistake of
genuine objection user 0%. Biometric devices can provide either high-speed recognition or
high security of the system. We have used three classes of fingerprints comparison
algorithms for the experiments. The algorithm with minutiae (individual points) is the
first class. The second is the correlation analysis. The latter ones are hybrid methods.
The most common, due to the simplicity and speed of work, is the method of
comparison by particular points - endpoints of papillary lines and dots of duplication of
papillary lines of minutiae. The description of minutiae was entirely carried out thoroughly
in [
The classification of minutiae is shown below. [
1. Ridge endings. It is a ridge that is located between two almost parallel crests. Ridge ending is the point where the ridge ends suddenly and does not appear again (Figure 4.a).
2. Bifurcations or Fork. A ridge divides the left side of the papillary lines into specific lengths and forms two parallel lines (Fig. 4.b), which sometimes can form a trifurcation (Fig. 4.d and 4.h).
3. Eye or enclosure. These are ellipsoid minutiae, which are formed by a ridge that branch out only for merge or approaching one crest, leaving space within the ridge. The enclosure may be small or large. (Fig. 4.c).
4. Convergence or convergent fork. This is similar to bifurcation, but with reciprocal or mirror image. It is formed by two parallel ridges. (Fig. 4.e).
5. Interjunction or Bridge minutiae. A joint between two parallel crests with a short diagonal of the ridge, which encounters the ridges in a very sharp angle (Fig. 4.f).
6. A fragment or short ridge. The ridge from the ends, which sharply ends, and have varying lengths. The fragment can be small or large (Fig. 4.g).
7. Hook or spur. It is formed at the vertebrae when the ridge divides into two parts (Fig. 4.h). One bifurcation ridge continues further, and another split is added to the ridge, as an appendage of the spine with a certain angle of inclination. The hook can be ascending crochet, and a descendant, hook, right crochet and left hook (Fig. 4.i and 4.j).
8. Return minutiae. One ridge suddenly turns back and forms a rounded loop. (Fig. 4.k). 9. Interruption. Interruptions formes between two crests that are interrupted, suddenly deviating, forming two ridges that end with a furrow between them (Fig. 4.l). 10. Crossovers are formed when two ridges cross each other (Fig. 4.m). 11. Dot is a tiny ridge that is usually found in the middle of the interruption, either delta or between the two ridges (Fig. 4.n).
12. Dotted ridge. This is a ridge that is created by dots (Fig. 4.o) [
The algorithm contains the steps of reading the imprint by the optical system,
recording to the image buffer, transferring to a convolution buffer, comparing data with
the template database, and deciding whether or not to identify. Because of the large
volume, we tried to encode biometric images using the RSA algorithm from [
The Adafruit fingerprint sensor is used as an optical sensor. The component
generates code for the Python programming languages. Programming was carried out on
Python in the PyCharm environment [
The comparison was carried out using the SSIM metric [
Figure 5 shows an analysis of the comparison for formula (1) of the originality of the prints and three different attempts to falsify and distort the original fingerprint and attempt to connect to the system. The experiments were carried out for three falsifications, and the results were displayed, the system did not allow any attempt and identified it as a fake.
Assessment of the local frequency of the ridges. If minutiae wasn't detected locally, the brightness levels along the crests could be modeled as a sinusoidal wave along the normal to the spine orientation. Fig. 6 shows the percentage display of the original fingerprint and failure in the system in 14 experiments.
This method was used in the construction of intelligent decision support systems
[
An identification system of biometric images using passive elements has been developed. The article analyzed the types of attacks affecting the identification system. Conclusions are made about the vulnerability of the system to attacks on it. Classification of minucius for biometric images has been carried out. We proposed to use of wave algorithm for skeletonization. Besides, experimental studies of the identification of biometric fingerprint are presented.