Fig 13 - uploaded by Arun Ross
Content may be subject to copyright.
Source publication
Most fingerprint-based biometric systems store the minutiae template of a user in the database. It has been traditionally assumed that the minutiae template of a user does not reveal any information about the original fingerprint. In this paper, we challenge this notion and show that three levels of information about the parent fingerprint can be e...
Context in source publication
Context 1
... Fig. 7 summarizes the main steps in the proposed classification algorithm. A visual analysis of the fingerprint ridge patterns of various classes reveals that they have almost the same ridge structure in the base and marginal areas, as shown in Fig. 8. However, it is the irregularities in the vicinity of the core region that are significant for classification (such as the circular ridge pattern in the case of whorls or the curving back of ridges in loops). In order to select these “salient” minutiae, we attempt to detect a registration point ð R 0 Þ using the Hough transform [3]. The ridges around the core point have a high curvature and form a nearly circular pattern. Accordingly, the orientations of the minutiae in such regions define a nearly circular pattern. Consider a circle ð x À x 0 Þ 2 þ ð y À y 0 Þ 2 1⁄4 r 2 defined by a group of minutiae, as shown in Fig. 9a. Here, x 0 and y 0 are the coordinates of the center of the circle and r is its radius. Our goal here is to detect the center of the circle. It can be observed that the minutiae depicted in Fig. 9a have orientations almost tangential to the circum- ference of the circle. Using this property, for each minutiae m we traverse a line L that is perpendicular to its orientation . This line is viewed as a set of discrete points, as shown in Fig. 9b. These points correspond to the probable center points of the circle. The radius r of each of these circles is the distance between the minutiae and the corresponding center points since the minutia m lies on their circumferences. An accumulator in 3D Hough space corresponding to the center ð x; y Þ and radius r is used to detect the center. For a circular minutiae pattern, there will be a well-pronounced peak in the Hough parameter space corresponding to the center ð x 0 ; y 0 Þ (point marked with red “x” in Fig. 9a). This is the registration point characterized by significant minutiae activity around it. Since the ridge structure in fingerprints is not exactly circular, we use lines that are nearly perpendicular ðÆ 30 Þ to the orientation , as shown in Fig. 9c. The results of the Hough transform using the estimated orientation maps and the minutiae points are shown in Fig. 10. The purpose of this exercise is not to detect the core of the fingerprint; rather, it is to detect a registration point that can be used to extract the “salient” minutiae. Only the minutiae located in a 300  300 pixels region about R 0 are used for classification (Fig. 11). A feature vector is next extracted from the set of salient minutiae identified in the previous stage. The features in this vector capture various properties of the minutiae such as the relationship between minutiae location and orientation, the clustering property of minutiae, the relationship between minutiae pairs, etc. The features we have designed in this regard are invariant to the rotation and translation of fingerprint images. The 11-dimensional feature vector F 1⁄4 f F 1 ; F 2 ; Á Á Á ; F 11 g is constructed as follows. The orientations represented by the minutiae vary across the four classes. For instance, a whorl fingerprint has at least one ridge which traverses a 360 closed path in the central region of the fingerprint. Thus, the orientations of these minutiae range from 0-360 degrees. On the other hand, the minutiae orientations of arches have only two dominant directions. In order to understand the distribution of minutiae orientations for each class, we examine the rose plots of minutiae orientations (Fig. 12). The range of dominant minutiae orientations is captured by counting the number of empty bins in the rose plot (feature F 1 ). This feature effectively discriminates between whorl and arch types, though it may not discriminate between the other classes. Feature F 2 is used to denote the variance in minutiae orientations present in the template. Minutiae pairs are fundamental units for representing variations in fingerprints [32]. The properties of neighboring minutiae change across classes. For instance, the neighboring minutiae in the central region of W have large orientation differences, whereas minutiae neighbors in A have similar orientations. The correlation between spatial location and orientation of minutiae pairs can be examined by estimating the joint distribution of R and È , where R is the distance between two minutiae and È is the difference in their orientation (as shown in (4) to (7)). Let P ð R; È Þ denote the probability of observing a minutiae pair which are separated by a distance R and with difference in orientation È . Then, F 3 is the number of minutiae pairs that are spatially compact and have almost similar orientations; F 4 is the number of pairs that are spatially compact but have a large orientation difference; F 5 is the number of minutiae pairs that are spatially separated and have similar orientations; F 6 is the number of minutiae pairs that are spatially separated but have large differences in orientation. Fig. 13 illustrates these four types of minutiae pairs. In our experiments with the NIST-4 database, we have set R 1 1⁄4 60 pixels , R 2 1⁄4 180 pixels , È 1 1⁄4 30 , and È 2 1⁄4 180 . It is observed that these features are significantly different for the four fingerprint classes. For arches, F 3 takes relatively larger values compared to the other three features, whereas, for whorls, F 4 is typically the largest. Minutiae tend to cluster in certain regions of the fingerprint. For example, minutiae activity increases in the core and delta regions of a fingerprint [21], [32], [8]. Variations in ridge flow seem to contribute to a high incidence of minutiae points. In order to capture these variations across fingerprint classes, we compute the minutiae density in circular regions representing a radii of 50 pixels. Feature F 7 is defined to be the maximum minutiae density corresponding to a particular fingerprint template. The value of F 7 is relatively high for whorls and small for arches. Feature F 8 is the maximum variance in minutiae orientations observed across all the circular regions considered. Visually (Fig. 1), it is apparent that features F 1 to F 8 can possibly distinguish classes A and W , but are not sufficient for reliably resolving ambiguity between classes ( L , R ), ( A , L , R ), and ( L , R , W ). It is necessary to include information about global ridge pattern in conjunction with the local minutiae properties to address this issue. To capture the global ridge structure of the fingerprint, we define geometric kernels which model the shape of the fingerprint around the core region for the W , L , and R classes. In a left loop, the ridges in the core region form a loop by recurving to the left side of the fingerprint. This is captured by a kernel constructed using two semi-ellipses corresponding to the concave and convex portions of the loop. The kernel for R is merely a mirror image of the L kernel. The circular ridge structure of W is captured using a simple circular kernel. Since the marginal area of every fingerprint has arch-like characteristics (Fig. 8), we do not define a separate kernel for the A class as it would align well with fingerprints of all the other classes. See Fig. 14. In order to determine a goodness-of-fit for these kernels, we use a model-based scheme by modifying the hierarchical kernel fitting approach proposed by Jain and Minut [18]. In this approach, fingerprint classification is achieved by finding the kernel that best fits the flow field (i.e, orientation field) of a given fingerprint template. Consider V to be a smooth vector field defined over some region in the plane R 2 and let be its argument. Let be a circular kernel curve in R 2 , as shown in Fig. 15. Let _ be the tangent to and be its argument. Consider a point t 1⁄4 ð x ð t Þ ; y ð t ÞÞ present on the kernel . An energy functional capturing the difference between the direction of _ and that of the vector field V at point is defined by To determine how well each kernel fits the estimated orientation map of the fingerprint, energy values (6) for various discrete points on the kernel are computed. The average of these values is used to determine the goodness of the fit. A lower energy average means a better fitting kernel. For each estimated orientation map (from minutiae information), we find three values for the energy functionals corresponding to the L , R , and W kernels. These functionals form features ( F 9 , F 10 , F 11 ), respectively. For an orientation map corresponding to W , it is expected that the minimum of the three features will be F 11 , whereas, for L and R , the minimum will correspond to F 9 and F 10 , respectively. We observed that, due to the inherent similarity in ridge structure of loops and arches, both the L and R kernels fit well for fingerprints belonging to class A . Thus, F 9 and F 10 are similar-valued for arches. This property is useful for resolving the ambiguity between loops and arches. The class-specific kernels are defined with respect to the registration point, R 0 , obtained using the Hough transform, i.e., for the W kernel, R 0 is the center of the circle, whereas, for the L and R kernels, R 0 is the focus of the elliptical kernel. The rotation and translation of fingerprints are taken into account by subjecting these kernels to certain predefined transformations prior to applying them. For W , the radius of the kernel is varied between 100 to 160 pixels. For L and R , the semi-major axis is varied between 120 to 180, while the semi-minor axis is varied between 60 to 100 pixels. The angle that the ellipse subtends with the horizontal is varied between À 10 to 10 degrees. Further, the kernels are moved in a 20  20 window around R 0 . The features ( F 9 , F 10 , F 11 ) correspond to the minimum of the energy functional obtained across all transformations for each kernel. A K -nearest neighbor classifier employing the Manhattan distance was used to ...
Similar publications
The preservation of musical works produced in the past requires their digitalization and transformation into a machine-readable format. The processing of handwritten musical scores by computers remains far from ideal. One of the fundamental stages to carry out this task is the staff line detection. We investigate a general-purpose, knowledge-free m...
Two-dimensional echocardiography (2D-echo) allows the evaluation of cardiac structures and their movements. A wide range of clinical diagnoses are based on the performance of the left ventricle. The evaluation of myocardial function is typically performed by manual segmentation of the ventricular cavity in a series of dynamic images. This process i...
We focus on the problem of developing a coupled statistical model that can be used to recover facial shape from brightness images of faces. We study three alternative representations for facial shape. These are the surface height function, the surface gradient, and a Fourier basis representation. We jointly capture variations in intensity and the s...
Diffuse optical imaging is an emerging medical imaging modality based on near-infrared and visible red light. The method can be used for imaging activations in the human brain. In this study, a deformable probabilistic atlas of the distribution of tissue types within the term neonatal head was created based on MR images. The use of anatomical prior...
A discriminant formulation of top-down visual saliency, intrinsically connected to the recognition problem, is proposed. The new formulation is shown to be closely related to a number of classical principles for the organization of perceptual systems, including infomax, inference by detection of suspicious coincidences, classification with minimal...
Citations
... 4. Cross-matching can be used to track individuals without their consent. To protect biometric template databases from being compromised for supporting widespread use of biometric based authentication is an important research challenge and a critical step in the successful implementation of a BIMS [8]. Every effective BIMS must satisfy some basic security and privacy requirements: ...
Rapid development of automation in the day-to-day life activity marks up the need of securing bio-metric template and the privacy of rightful owner. Minutiae-based matching is the most popular in the fingerprint recognition system, which greatly suffers from non-linear distortion like translation and rotation. To deal with linear distortion most of the technique proposed in the literature depends upon a reference or singular point. The paper proposes a binary template generation technique which applies an unsupervised clustering technique without fixing the number of clusters. Instead of position and orientation of the minutiae points the cardinality of the clusters are stored and converted into binary template. No spatial pattern information about the fingerprint is stored in the template to protect it from spoofing and information leakage. By the help of modified Radial Basis Function Network (mRBFN) with robust and efficient matching technique the generated templates are matched for authentication. We use MCYT dataset for training the mRBFN. The efficiency of the proposed scheme is evaluated on FVC 2000, FVC 2002 and FVC 2004 dataset.
... Reconstructing fingerprints from the minutiae templates or the inversion of fingerprint templates was initially addressed in 2001 [35]. However, intensive research in this area started in 2007 by [36,37]. Regarding the model-based approaches, the process is technically similar to fingerprint synthesis with the only difference being that singularity points and the orientation map are estimated from minutiae and not vice versa. ...
... The final rendering step enhances realism. In [37], a fingerprint image is generated from a skeleton image, which is reconstructed from minutiae. The process involves estimating an orientation map using minutiae triplets and drawing streamlines starting from minutiae and border points. ...
The datasets of synthetic biometric samples are created having in mind two major objectives: bypassing privacy concerns and compensating for missing sample variability in datasets of real biometric samples. If the purpose of generating samples is the evaluation of biometric systems, the foremost challenge is to generate so-called mated impressions—different fingerprints of the same finger. Note that for fingerprints, the finger’s identity is given by the co-location of minutiae points. The other challenge is to ensure the realism of generated samples. We solve both challenges by reconstructing fingerprints from pseudo-random minutiae making use of the pix2pix network. For controlling the identity of mated impressions, we derive the locations and orientations of minutiae from randomly created non-realistic synthetic fingerprints and slightly modify them in an identity-preserving way. Our previously trained pix2pix models reconstruct fingerprint images from minutiae maps, ensuring that the realistic appearance is transferred from training to synthetic samples. The main contribution of this work lies in creating and making public two synthetic fingerprint datasets of 500 virtual subjects with 8 fingers each and 10 impressions per finger, totaling 40,000 samples in each dataset. Our synthetic datasets are designed to possess characteristics of real biometric datasets. Thus, we believe they can be applied for the privacy-friendly testing of fingerprint recognition systems. In our evaluation, we use NFIQ2 for approving the visual quality and Verifinger SDK for measuring the reconstruction success.
... It is also worth observing that, for an attacker, collecting the templates stored in a database could be as effective as acquiring the original biometric data. In fact, it has been shown for several biometric identifiers that the original biometric traits can be adequately reconstructed from their representations, namely the templates [28], and that such reverse process can be also performed when the employed features are obtained through the use of neural networks [20]. ...
Representations of biometric traits to be used in automatic recognition systems are typically learned with the goal of obtaining significant discriminative capabilities, that is, generating features that are notably different when produced by traits of different subjects, while maintaining an appropriate consistency for a given user. Nonetheless, discriminability is not the only desirable property of a biometric representation. For instance, the mutual independence of the coefficients in the employed templates is a valuable property when designing biometric template protection schemes. In fact, managing representations with independent coefficients allows to maximize the achievable security. In this paper we propose different learning strategies to obtain biometric representations with the property of statistical independence among coefficients, while preserving discriminability. In order to achieve this goal, different strategies are employed to train convolutional autoencoders. As a proof of concept, the effectiveness of the proposed approaches is tested by considering biometric recognition systems using both finger-vein and palm-vein patterns.
... To mitigate these security issues, it is crucial to implement robust security protocols, including encryption and access controls, to protect sensitive data. Regular auditing and monitoring of human-machine interactions can help identify and prevent malicious activities [11]. Additionally, ensuring diversity and inclusivity in the development and training of AI algorithms can help mitigate algorithmic bias and promote fairness in identity verification processes. ...
The Human-Machine Identity Verification System based on Deep Learning offers a robust and automated approach to identity verification, leveraging the power of deep learning algorithms to enhance accuracy and security. This paper focused on the biometric-based authentical scheme with Biometric Recognition for the Huma-Machinary Identification System. The proposed model is stated as the Two-Factor Biometric Authentication Deep Learning (TBAuthDL). The proposed TBAuthDL model uses the iris and fingerprint biometric data for authentication. TBAuthDL uses the Weighted Hashing Cryptographic (WHC) model for the data security. The TBAuthDL model computes the hashing factors and biometric details of the person with WHC and updates to the TBAuthDL. Upon the verification of the details of the assessment is verified in the Human-Machinary identity. The simulation analysis of TBAuthDL model achieves a higher accuracy of 99% with a minimal error rate of 1% which is significantly higher than the existing techniques. The performance also minimizes the computation and processing time with reduced complexity.
... In [32], it is shown how ridge orientations can be modeled by fitting Legendre polynomials. [30] Realistic fingerprint synthesis based on [31] Statistical ridge pattern model Ram et al. '10 [32] Estimation of ridge orientation Modelling ridges by Legendre polynomials Johnson et al. '13 [33] Texture mapping from real to master fingerprint Modeling of ridge texture & valley noise features Drahansky '19 [34] Simulation of sensor, environment, user and Modeling of synthesis states based on Petri net finger conditions, skin diseases Priesnitz et al. '22 [35] Synthesis of contactless fingerprints Simulation of capturing process, subject characteristics, and environmental influences Fingerprint reconstruction from minutiae Cappelli et al. '07 [36] Fingerprint reconstruction Orientation from zero-pole model Ross et al. '07 [37] Partial fingerprints reconstruction Orientation from minutiae triplets Feng and Jain '11 [38] Fingerprint reconstruction Orientation from AM-FM model Li and Kot '12 [39] Fingerprint reconstruction Orientation from AM-FM model Cao and Jain '15 [40] Idealistic ridge patterns (no noise) Patch dictionaries for orient. & continuous phase Face synthesis Sirovich and Kirby '87 [41] Face as a linear combination of eigenpictures Principal component analysis (PCA) for faces Applied to face recognition in [42] -Eigenfaces Software tools: EVOFIT, E-Fit V Cootes '01 [43] Active Appearance Model PCA on texture + PCA on facial landmarks Blanz and Vetter '99 [44] Morphable 3D Face Model PCA on texture + PCA on 3D face model Software Tool: FaceGen Kähler et al. '02 [45] Deformable head model with anatomical Modeling musculature and skeletal structure structure Iris synthesis Lefohn et al. '03 [46] Ocularist's model Stacking layers: stroma, collarette, sphincter, etc. Cui et al. '04 [47] Iris synthesis PCA + super-resolution Makthal and Ross '05 [48] Iris texture synthesis Markov random field (MRF) model Shah and Ross '06 [49] Realistic iris synthesis MRF model for background texture + embedding radial and concentric furrows, collarette and crypts Zuo et al. '07 [50] Compilation of large-scale synthetic iris datasets Anatomy based model Iris reconstruction from binary IrisCode Venugopalan et al. '11 [51] Iris reconstruction Deterministic approach based on Gabor filtering Galbally et al. '13 [52] Iris reconstruction Probabilistic approach based on genetic algorithms Synthesis of vascular samples Crisan et al. '08 [53] Realistic hand vein samples Vein nodes are connected by randomly driven vessel growth and imaging variation modelling Hillerström et al. '14 [54] Realistic finger vein samples Vein nodes are connected by randomly driven vessel growth and imaging variation modelling Fiorini et al. '14 [55] Realistic retina (fundus) samples Patch-based approach to create texture Bonaldi et al. '16 [56] Active shapes and Kalman filtering for vessels Das et al. '17 [57] Realistic sclera patterns Non-parametric texture synthesis ...
... Inversion of fingerprint templates was first addressed by Hill back in 2001 [59]. However, the intensive research started in 2007 with the studies of Cappelli et al. [36] and Ross et al. [37]. In contrast to fingerprint drawing from scratch, the locations of singular points are derived from the given minutiae and further on the orientation map. ...
... Ridge frequency are also adjusted to minutiae. Apart from the zeropole model [31], ridge patterns can be estimated using the VOLUME 10, 2022 minutiae triplet model [37] or the AM-FM model [38], [39] which makes use of eight neighbouring minutiae. Reconstruction of ridges based on patch dictionaries, proposed in [40], allows for idealistic ridge patterns. ...
Synthetic biometric samples are created with an ultimate goal of getting around privacy concerns, mitigating biases in biometric datasets, and reducing the sample acquisition effort to enable large-scale evaluations. The recent breakthrough in the development of neural generative models shifted the focus from image synthesis by mathematical modeling of biometric modalities to data-driven image generation. This paradigm shift on the one hand greatly improves the realism of synthetic biometric samples and therefore enables new use cases, but on the other hand new challenges and concerns arise. Despite their realism, synthetic samples have to be checked for appropriateness for the tasks they are intended which includes new quality metrics. Focusing on sample images of fingerprint, face, iris and vascular patterns, we highlight the benefits of using synthetic samples, review the use cases, and summarize and categorize the most prominent studies on synthetic biometrics aiming at showing recent progress and the direction of future research.
... These models are known as multi-modal methods and are particularly challenging in real life because a framework has to be developed that can handle multiple variations such as subject position and view obstructions [68,69]. Building architectures of deep learning models for person Re-ID is a very time-consuming task, so some researchers are using neural architecture search methods (NAS) to automate the process of architecture engineering [70,71]. The most crucial challenge in NAS methods is that there is no guarantee of how appropriate the CNN would be after NAS has chosen it. ...
Survey/review study A Review of Techniques on Gait-Based Person Re-Identification Babak Rahi 1,*, Maozhen Li 1, and Man Qi 2 1 Department of Electronics and Computer Engineering, Brunel University London, Uxbridge, Middlesex, UB8 3PH, United Kingdom 2 The School of Engineering, University of Warwick, Coventry CV4 7AL, United Kingdom * Correspondence: babak.h.rahi@hotmail.com Received: 16 October 2022 Accepted: 14 December 2022 Published: Abstract: Person re-identification at a distance across multiple non-overlapping cameras has been an active research area for years. In the past ten years, short-term Person re-identification techniques have made great strides in accuracy using only appearance features in limited environments. However, massive intra-class variations and inter-class confusion limit their ability to be used in practical applications. Moreover, appearance consistency can only be assumed in a short time span from one camera to the other. Since the holistic appearance will change drastically over days and weeks, the technique, as mentioned above, will be ineffective. Practical applications usually require a long-term solution in which the subject's appearance and clothing might have changed after the elapse of a significant period. Facing these problems, soft biometric features such as Gait has stirred much interest in the past years. Nevertheless, even Gait can vary with illness, ageing and emotional states, walking surfaces, shoe types, clothes types, carried objects (by the subject) and even environment clutters. Therefore, Gait is considered as a temporal cue that could provide biometric motion information. On the other hand, the shape of the human body could be viewed as a spatial signal which can produce valuable information. So extracting discriminative features from both spatial and temporal domains would benefit this research. This article examines the main approaches used in gait analysis for re-identification over the past decade. We identify several relevant dimensions of the problem and provide a taxonomic analysis of current research. We conclude by reviewing the performance levels achievable with current technology and providing a perspective on the most challenging and promising research directions.
... The corresponding relationship between minutiae is established by considering the similarity of minutiae descriptors, and then fingerprint comparison is performed. To improve the identification performance, many papers combine fingerprint minutiae points with other key information [30,31] as the feature representation of fingerprints, and then measure the similarity between them. Lee et al. [32] add the ridge feature to the traditional detail feature identification, and the final identification score is the combination of the two identification stages. ...
Fingerprints are the most widely used of all biological characteristics in public safety and forensic identification. However, fingerprint images extracted from the crime scene are incomplete. On the one hand, due to the lack of effective area in partial fingerprint images, the extracted features are insufficient. On the other hand, a broken ridge may lead to a large number of false feature points, which affect the accuracy of fingerprint recognition. Existing fingerprint identification methods are not ideal for partial fingerprint identification. To overcome these problems, this paper proposes an attention-based partial fingerprint identification model named APFI. Firstly, the algorithm utilizes the residual network (ResNet) for feature descriptor extraction, which generates a representation of spatial information on fingerprint expression. Secondly, the channel attention module is inserted into the proposed model to obtain more accurate fingerprint feature information from the residual block. Then, to improve the identification accuracy of partial fingerprints, the angular distance between features is used to calculate the similarity of fingerprints. Finally, the proposed model is trained and validated on a home-made partial fingerprint image dataset. Experiments on the home-made fingerprint datasets and the NIST-SD4 datasets show that the partial fingerprint identification method proposed in this paper has higher identification accuracy than other state-of-the-art methods.
... The fact that fingerprint impression can be reconstructed using the precise information of minutiae locations [10,14,34] gave rise to the idea of securing raw fingerprint data. This is ensured by biometric template protection schemes (BTPS). ...
The deployment of biometrics spans over almost every application that incorporates user authentication. Fingerprint biometrics in particular is by far the most popular choice in biometric authentication systems. However, vulnerabilities such as database compromise, fingerprint reconstruction, and identity fraud have made it crucial to develop highly robust and privacy-preserving authentication systems. Although there have been considerable efforts in the area, security gaps such as dependency on singular point, scale variability and low user recognition performance limit the real-time application of these techniques. In this paper, we address these performance issues and put forth a novel technique for fingerprint template security. The secure templates generated by the proposed technique are computed using many-to-one mapped fully quantized features. A set of rotation and translation independent features are derived using minutiae triplets which makes the proposed secure fingerprint template an alignment-free user template. Each feature type is rotation and translation invariant making the templates free for scale anomalies. The technique essentially hides the true feature information of a fingerprint with the help of a non-invertible transformation function. This enables user identity verification in the transformed domain itself. An extensive result analysis using five fingerprint databases obtained from FVC2002 and FVC2004 has been carried out. The proposed technique delivers a superior performance compared to other noteworthy contributions and therefore proves to be fit for use in various applications involving fingerprint-based identity verification mechanisms.
... However, potential risk associated with biometric recognition is that users are obliged to reveal their biometric data, which is a highly sensitive piece of personal information. Use of biometrics to ensure security might tamper user's privacy [3,43,44]. Biometric recognition system which ensures the security of biometric templates along with acceptable matching performance has paramount importance in this digital era [16,18]. ...
The idea of cancelable biometrics was proposed to ensure the privacy and security of user’s original biometric data, which is collected for recognition by storing only transformed template in the database after applying irreversible operations. State of the art cancelable schemes use either biometric salting or non-linear transformations for generating protected template. However, biometric salting techniques suffer from the threat of unauthorized access to user specific key; whereas non-linear transformation based schemes have the inherent issue of performance degradation due to information loss. In order to overcome these limitations of the existing techniques, this work proposes a patch level distortion based cancelable scheme which can be applied for multiple modalities. In the proposed method, biometric images are divided into Voronoi patches and log Gabor feature vector of each patch is distorted by combining it with a function of difference vectors of other patches. Proposed scheme is able to eliminate the need of security critical user specific key and many to one transformations. Irreversibility is assured as the function of difference vectors of patches used as noise in distortion process is random enough and unpredictable for an intruder. Cancelability and non-linkability can be ensured by changing the number and location of seed points. Superior performance compared to state of the art techniques is obtained in distorted domain when experiments were conducted on CASIA face database.
... Despite the fact that the usage of biometrics has made a remarkable breakthrough in the area of security, its unprotected use can open the flood gates to significant security and privacy concerns. One of the considerable risks in this regard is the permanent loss of the user's identity [9,12,35]. To mitigate these challenges, biometric template protection schemes have been employed. ...
Authentication systems have evolved immensely ever since the use of biometrics in the field of security has started. Nonetheless, extensive usage of biometric systems has also resulted in growing fear of identity theft as storing biometric user templates in databases opens up severe security concerns. Consequently, the scope of biometrics invariably depends on the ability of the system to manifest security and robustness against biometric identity theft along with achieving acceptable recognition performance. This paper proposes a user template protection technique to secure a fingerprint based user template used in a biometric authentication system. It is based on the fingerprint shell, which computes alignment-free, singular point independent, and non-invertible user templates. The secure user template generated by the proposed technique comprises fingerprint shell curves computed using the transformed pair-polar structure of the minutia points. The proposed technique has been evaluated on five fingerprint databases of Fingerprint Verification Competitions, namely, FVC2002, FVC2004, and FVC2006. The effectiveness of the technique is analyzed in terms of revocability, diversity, security, and recognition performance. The comparison of results with that of the existing techniques demonstrates the robustness and efficacy of the proposed technique.
