Showing papers by "Christoph Busch published in 2012"
TL;DR: This survey categorise and summarise approaches to ear detection and recognition in 2D and 3D images, and provides an outlook over possible future research in the field of ear recognition, in the context of smart surveillance and forensic image analysis, which the authors consider to be the most important application ofEar recognition characteristic in the near future.
Abstract: The possibility of identifying people by the shape of their outer ear was first discovered by the French criminologist Bertillon, and refined by the American police officer Iannarelli, who proposed a first ear recognition system based on only seven features. The detailed structure of the ear is not only unique, but also permanent, as the appearance of the ear does not change over the course of a human life. Additionally, the acquisition of ear images does not necessarily require a person's cooperation but is nevertheless considered to be non-intrusive by most people. Owing to these qualities, the interest in ear recognition systems has grown significantly in recent years. In this survey, the authors categorise and summarise approaches to ear detection and recognition in 2D and 3D images. Then, they provide an outlook over possible future research in the field of ear recognition, in the context of smart surveillance and forensic image analysis, which they consider to be the most important application of ear recognition characteristic in the near future.
214 citations
18 Jul 2012
TL;DR: This paper extracts several features from the gait data and uses the k-Nearest Neighbour algorithm for classification and shows that this algorithm yields a better biometric performance than the machine learning algorithms the authors previously used for classification, namely Hidden Markov Models and Support Vector Machines.
Abstract: Accelerometer-based biometric gait recognition offers a convenient way to authenticate users on their mobile devices. Modern smartphones contain in-built accelerometers which can be used as sensors to acquire the necessary data while the subjects are walking. Hence, no additional costs for special sensors are imposed to the user. In this publication we extract several features from the gait data and use the k-Nearest Neighbour algorithm for classification. We show that this algorithm yields a better biometric performance than the machine learning algorithms we previously used for classification, namely Hidden Markov Models and Support Vector Machines. We implemented the presented method on a smartphone and demonstrate that it is efficient enough to be applied in practice.
153 citations
06 Aug 2012
TL;DR: This paper presents a selection of well-defined criteria and some metrics that are compliant with the reference architecture for template protection as defined in the recently adopted standard ISO/IEC 24745 (2011), which is applicable to nearly all known BTP methods.
Abstract: Traditional criteria used in biometric performance evaluation do not cover all the performance aspects of biometric template protection (BTP) and the lack of well-defined metrics inhibits the proper evaluation of such methods. Previous work in the literature focuses, in general, on a limited set of criteria and methods. This paper provides the first holistic approach to the evaluation of biometric template protection that is able to cover a whole range of methods. We present a selection of well-defined criteria and some metrics that are compliant with the reference architecture for template protection as defined in the recently adopted standard ISO/IEC 24745 (2011), which is applicable to nearly all known BTP methods. The criteria have been grouped in three categories of performance: technical, protection, and operational.
79 citations
Proceedings Article•
27 Sep 2012TL;DR: The proposed authentication method is analyzed for feasibility and implemented in a prototype as application for the Android operating system, and a biometric database containing photos of the two test devices from 41 test subjects is created.
Abstract: This paper is concerned with the authentication of people on smartphones using fingerphoto recognition. In this work, fingerphotos are captured with the built-in camera of the smartphone. The proposed authentication method is analyzed for feasibility and implemented in a prototype as application for the Android operating system. Algorithms for the capture process are developed to ensure a minimum of quality of the captured photos to enable a reliable fingerphoto recognition. Several methods for preprocessing of the captured samples are analyzed and performant solutions to evaluate the photos are developed to enhance the recognition rates. This is achieved by evaluating a wide range of different parameters and configurations of the algorithms as well as various combinations of preprocessing chains for the captured samples. The operations for preprocessing are selected with respect to their computational effort to guarantee that they can be executed on a smartphone with limited computation and memory capacity. The developed prototype is evaluated in user tests with two different smartphones. Additionally, a biometric database containing photos of the two test devices from 41 test subjects is created. These fingerphotos are used to evaluate and optimize the procedures.
67 citations
28 Nov 2012
TL;DR: This bookchapter provides a comprehensive overview of biometric template protection schemes which provide provable security/ privacy, and achieve practical recognition rates have remained elusive, even on small datasets.
Abstract: The term biometrics refers to “automated recognition of individuals based on their behavioral and biological characteristics” (ISO/IEC JTC1 SC37). Several physiological (static) as well as behavioral (non-static) biometric characteristics have been exploited (Jain, Flynn & Ross, 2008) such as fingerprints, iris, face, hand, voice, gait, keystroke dynamics, etc., depending on distinct types of applications (see Figure 1). Biometric traits are acquired applying adequate sensors and distinctive feature extractors are utilized in order to generate a biometric template (reference data) in the enrollment process. During verification (authentication process) or identification (identification can be handled as a sequence of biometric comparisons against the enrollment records in a reference databse) the system processes another biometric measurement from which an according template is extracted and compared against the stored template(s) yielding acceptance/ rejection or hit/ no-hit, respectively. The presented work is motivated by very recent advances in the fields of multi-biometric recognition (Ross et al., 2006) and biometric template protection (Rathgeb & Uhl, 2011). Automatic recognition systems based on a single biometric indicator often have to contend with unacceptable error rates (Ross & Jain, 2003). Multi-biometric systems have improved the accuracy and reliability of biometric systems (Ross et al., 2006). Biometric vendors are already deploying multi-biometric systems (e.g. fingerprint and finger vein by SAFRAN Morpho1) and multi-biometric recognition is performed on large-scale datasets (e.g. within the Aadhaar project (Unique Identification Authority of India, 2012) by the Unique Identification Authority of India (UIDAI)). However, security of multi-biometric templates is especially crucial as they contain information regarding multiple traits of the same subject (Nagar et al., 2012). The leakage of any kind of template information to unauthorized individuals constitutes serious security and privacy risks, e.g. permanent tracking of subjects without consent (Ratha et al., 2001) or reconstruction of original biometric traits (e.g. fingerprints (Cappelli et al., 2007) or iris textures (Venugopalan & Savvides, 2011)) might become a realistic threat. Therefore, biometric template protection technologies have been developed in order to protect privacy and integrity of stored biometric data. However, so far, template protection schemes which provide provable security/ privacy, and achieve practical recognition rates have remained elusive, even on small datasets. This bookchapter provides a comprehensive overview of
56 citations
TL;DR: The proposed solution based on spectral minutiae is evaluated against other comparison strategies on three different datasets of wrist and palm dorsal vein samples and shows a competitive biometric performance while producing features that are compatible with state-of-the-art template protection systems.
Abstract: Similar to biometric fingerprint recognition, characteristic minutiae points - here end and branch points - can be extracted from skeletonised vein images to distinguish individuals. An approach to extract those vein minutiae and to transform them into a fixed-length, translation and scale invariant representation where rotations can be easily compensated is presented in this study. The proposed solution based on spectral minutiae is evaluated against other comparison strategies on three different datasets of wrist and palm dorsal vein samples. The authors- analysis shows a competitive biometric performance while producing features that are compatible with state-of-the-art template protection systems. In addition, a modified and more distinctive, but not transform or rotation invariant, representation is proposed and evaluated.
52 citations
06 Aug 2012
TL;DR: The proposed quality metric is based on Gabor filter responses and is evaluated against eight contemporary quality estimation methods on four datasets using sample utility derived from the separation of genuine and imposter distributions as benchmark and shows performance and consistency approaching that of the composite NFIQ quality assessment algorithm.
Abstract: Quality assessment of biometric fingerprint images is necessary to ensure high biometric performance in biometric recognition systems. We relate the quality of a fingerprint sample to the biometric performance to ensure an objective and performance oriented benchmark. The proposed quality metric is based on Gabor filter responses and is evaluated against eight contemporary quality estimation methods on four datasets using sample utility derived from the separation of genuine and imposter distributions as benchmark. The proposed metric shows performance and consistency approaching that of the composite NFIQ quality assessment algorithm and is thus a candidate for inclusion in a feature vector introducing the NFIQ 2.0 metric.
44 citations
TL;DR: A new chain code based feature en- coding method is proposed, using spatial and orientation properties of vein patterns, which is capable of dealing with noisy and unstable image skeletons.
24 citations
Proceedings Article•
11 Apr 2012
TL;DR: This approach realizes user authentication by applying the discrete wavelet transform (DWT) to acceleration signals obtained from mobile devices by observing the user's gait characteristic to serve both convenience and security needs at the same time.
Abstract: The ever growing number of mobile devices has turned the attention to security and usability. If a mobile device is lost or stolen this can lead to loss of personal information and the possibility of identity theft. People often tend not to use passwords which leads to lack of personal security mainly due to convenience and frequent use. This paper suggests to serve both convenience and security needs at the same time. Thus we suggest to observe the user's gait characteristic. Our approach realizes user authentication by applying the discrete wavelet transform (DWT) to acceleration signals obtained from mobile devices. Gait templates were constructed of Bark-frequency cepstral coefficients (BFCC) from the wavelet coefficients and these were arranged to train a support vector machine (SVM). A cross-day scenario demonstrates that the proposed approach shows competitive recognition performance, yielding 9.82% False Match Rate (FMR) at a False Non-Match Rate (FNMR) of 10.45%.
21 citations
06 Aug 2012
TL;DR: This work shows that distribution analysis is essential for security assessment of fuzzy commitment, and shows that with knowledge of the iris distribution secrets can be recovered with low complexity.
Abstract: Iris patterns contain rich discriminative information and can be efficiently encoded in a compact binary form. These nice properties allow smooth integration with the fuzzy commitment scheme. Instead of storing iris codes directly, a random secret can be derived such that user privacy can be preserved. Despite the successful implementation, the dependency existing in iris codes can strongly reduce the security of fuzzy commitment. This paper shows that the distribution of iris codes complies with the Markov model. Additionally, an algorithm retrieving secrets from the iris fuzzy commitment scheme is proposed. The experimental results show that with knowledge of the iris distribution secrets can be recovered with low complexity. This work shows that distribution analysis is essential for security assessment of fuzzy commitment. Ignoring the dependency of binary features can lead to overestimation of the security.
17 citations
TL;DR: This approach directly estimates the maximum ridge frequency orientation by the amplitude-frequency features of the Fast Fourier Transform and takes the frequency features' difference in two perpendicular orientations as a distinguishing feature for ridge-like patterns to distinguish the high-quality blocks from other low-quality ones or background area.
Abstract: We present in this paper a sample quality control approach for the case using a mobile phone's camera as a fingerprint
sensor for fingerprint recognition. Our approach directly estimates the maximum ridge frequency orientation by the
amplitude-frequency features of the Fast Fourier Transform and takes the frequency features' difference in two
perpendicular orientations as a distinguishing feature for ridge-like patterns. Then a decision criterion which combines
the frequency components' energy and ridge orientation features is used to determine if an image block should be
classified as high-quality fingerprint area or not. The number of such high-quality blocks can thus be used to indicate the
whole fingerprint sample's quality. Experiments show this approach's effectiveness in distinguishing the high-quality
blocks from other low-quality ones or background area. Mapping the quality metric to the sample utility as derived from
the the NIST minutiae extractor "mindtct" function is also given to verify the approach's quality prediction effectiveness.
Keywords: Fingerprint, quality assessment, mobile phone camera
18 Jul 2012
TL;DR: A new ear detection approach for 3D profile images based on surface curvature and semantic analysis of edge-patterns is presented, which is robust against rotation and scale.
Abstract: Although a number of different ear recognition techniques have been proposed, not much work has been done in the field of ear detection. In this work we present a new ear detection approach for 3D profile images based on surface curvature and semantic analysis of edge-patterns. The algorithm applies edge-based detection techniques, which are known from 2D approaches, to a 3D data model. As an additional result of the ear detection, the outline of the outer helix is found, which may serve as a basis for further feature extraction steps. As our method does not use a reference ear model, the detector does not need any previous training. Furthermore, the approach is robust against rotation and scale. Experiments using the 3D images from UND-J2 collection resulted in a detection rate of 95.65\%.
31 Dec 2012
TL;DR: A close look is taken at fuzzy commitment, which is an efficient and widely used template protection algorithm and rigorous assessment of an iris fuzzy commitment scheme using the information-theoretical metrics is demonstrated.
Abstract: Template protection techniques are important supplements to biometrics, which aim to improve system security and safeguard privacy of users. Their development brings a new challenge of privacy and security assessment especially for real systems. In the paper, we take a close look at fuzzy commitment, which is an efficient and widely used template protection algorithm and demonstrates rigorous assessment of an iris fuzzy commitment scheme using the information-theoretical metrics. For instance, a 56 bit long secret can be derived from iris codes. Instead of iris codes, its hash value is stored. However, due to the dependency of iris codes, the uncertainty of secrets reduces to 11.82 bits given protected templates. It confirms the empirical results that an adversary is able to retrieve the iris features from the protected templates with average number of attempts equal to 210.56 as shown in [1]. The poor security and privacy performance is caused by strong correlation of iris feature and unsuitable coding methods used in the algorithm. The quantitative measurement shown in this paper provides a reference guidance on evaluation of template protection in practice. It helps algorithm developers to show the security and privacy of template protection to end-users and to detect the weaknesses of the algorithms.
02 Jul 2012
TL;DR: This research enhances the segmentation by a cycle extraction method and shows that the fixed-length segmentation is the better approach as it achieves similar error rates while requiring a lower computational effort.
Abstract: When machine learning algorithms are used for accelerometer-based biometric gait recognition, the general approach is to divide the data into segments of a fixed time-length, extract features from the segments and use these feature vectors to train the classifiers and authenticate the subjects. In this research we enhance the segmentation by a cycle extraction method. A gait cycle contains data of two steps, and our technique assures that each segment contains the same amount of steps. These cycle-based segments are input to the feature extraction process. We apply Hidden Markov Models (HMMs) for classification and compare the results to previous ones obtained using segments of a fixed time-length. We show that the fixed-length segmentation is the better approach as it achieves similar error rates while requiring a lower computational effort.
18 Jul 2012
TL;DR: Whether deletion of identity references, as the simplest anonymization way, is appropriate for privacy protection is discussed, and a closer look at the identity attributes is taken and to which degree they achieve privacy for an entity based on the proposed criteria.
Abstract: We analyze the privacy implications of identity references in biometric databases in which either raw biometric samples or templates are stored. The analysis was inspired by the privacy requirement imposed by the Norwegian National DPA (Data Protection Act Privacy Advisor) on the fingerprint data collection for one of our research projects. The respective DPA approved maintenance of a fingerprint database but required deletion of personal information, e.g., name, gender, age, and email address which were collected and stored in a 'namelist' (XLS-file) as identity references and maintained on a separate hard-disk from the biometric references. We discuss in this paper whether deletion of identity references, as the simplest anonymization way, is appropriate for privacy protection. We take a closer look at the identity attributes and assess to which degree they achieve privacy for an entity based on the proposed criteria. We analyze different biometric database models against the given criteria to give privacy protection guidelines for the identity reference construction and maintenance in a biometric database. In addition, the possibility of joint protection of identity references and biometric references is also investigated in some proposed models suitable for the identity authentication scenario.
Book Chapter•
01 Jan 2012
TL;DR: This is the publisher's copy of the book chapter originally published in Sicily 2012: Sicherheit, Schutz and Zuverlassigkeit Beitrage der 6.
Abstract: This is the publisher's copy of the book chapter originally published in: Suri, N. & Waidner, M. (ed.) (2012) Sicherheit 2012: Sicherheit, Schutz und Zuverlassigkeit Beitrage der 6. Jahrestagung des Fachbereichs Sicherheit der Gesellschaft fur Informatik e.V. (GI); 7.–9. Marz 2012, Darmstadt: Gesellschaft fur Informatik.Reprinted with permission from Gesellschaft fur Informatik.
31 Dec 2012
TL;DR: The detection performance of the modified HCS detector is evaluated on two different datasets, one of them containing images n various poses, and the modified approach by Zhou et al. towards making it invariant to rotation by using a rotation symmetric, circular detection window is modified.
Abstract: In identity retrieval from crime scene images, the outer ear (auricle) has ever since been regarded as a valuable characteristic. Because of its unique and permanent shape, the auricle also attracted the attention of researches in the field of biometrics over the last years. Since then, numerous pattern recognition techniques have been applied to ear images but similarly to face recognition, rotation and pose still pose problems to ear recognition systems. One solution for this is 3D ear imaging. the segmentation of the ear, prior to the actual feature extraction step, however, remains an unsolved problem. In 2010 Zhou at al. have proposed a solution for ear detection in 3D images, which incorporates a nave classifier using Shape Index Histogram. Histograms of Categorized Shapes (HCS) is reported to be efficient and accurate, but has difficulties with rotations. In our work, we extend the performance measures provided by Zhou et al. by evaluating the detection rate of the HCS detector under more realistic conditions. This includes performance measures with ear images under pose variations. Secondly, we propose to modify the ear detection approach by Zhou et al. towards making it invariant to rotation by using a rotation symmetric, circular detection window. Shape index histograms are extracted at different radii in order to get overlapping subsets within the circle. The detection performance of the modified HCS detector is evaluated on two different datasets, one of them containing images n various poses.
19 Nov 2012
TL;DR: The objective was not only to increase the recognition rate but also to develop a new, fake resistant capture device, and methods for protection of the biometric template were researched and the second generation of the international standard ISO/IEC 19794-5:2011 was inspired by the project results.
Abstract: Biometric data have been integrated in all ICAO compliant passports, since the ICAO members started to implement the ePassport standard. The additional use of three-dimensional models promises significant performance enhancements for border control points. By combining the geometry- and texture-channel information of the face, 3D face recognition systems show an improved robustness while processing variations in poses and problematic lighting conditions when taking the photo. This even holds in a hybrid scenario, when a 3D face scan is compared to a 2D reference image. To assess the potential of three-dimensional face recognition, the 3D Face project was initiated. This paper outlines the approach and research results of this project: The objective was not only to increase the recognition rate but also to develop a new, fake resistant capture device. In addition, methods for protection of the biometric template were researched and the second generation of the international standard ISO/IEC 19794-5:2011 was inspired by the project results.
06 Aug 2012
TL;DR: A score-level fusion rule is presented which assigns match cases to different match voting layers and weighs them according to their ranking in posterior probability of a genuine match, which achieves an average EER over the database FVC2002 DB2_A under the token-stolen scenario.
Abstract: Minutiae vicinity [1] captures a fingerprint's local topological information among neighboring minutiae for biometric template protection. We extend in this paper the definition of a minutia vicinity to a general one, which deems the conventional definition as a special case. Under this generalized definition, multiple vicinities can be obtained via defining different radial distances from one central minutia. Then fusion of these multiple vicinities can be done for better biometric performance. We present a score-level fusion rule which assigns match cases to different match voting layers and weighs them according to their ranking in posterior probability of a genuine match. Experiments achieve an average EER=0.0143 (compared to EER=0.0252 in the single vicinity case [1]) over the database FVC2002 DB2_A under the token-stolen scenario i.e., using the same public transformation parameters when generating protected vicinities for comparison.
Proceedings Article•
27 Sep 2012TL;DR: The use of betti numbers to characterize fingerprint and iris images is discussed and the proposed method is compared against SIVV, a tool provided by NIST, to demonstrate the potential of the scheme.
Abstract: This paper discusses the use of betti numbers to characterize fingerprint and iris images. The goal is to automatically separate fingerprint images from non-fingerprint images; where non-fingerprint images of special interest are biometric samples which are not fingerprints. In this regard, an image is viewed as a triangulated point cloud and the topology associated with this construct is summarized using its first betti number - a number that indicates the number of distinct cycles in the triangulation associated to the particular image. This number is then compared against the first betti numbers of “n” prototype images in order to perform classification (“fingerprint” vs “non-fingerprint”). The proposed method is compared against SIVV (a tool provided by NIST). Experimental results on fingerprint and iris databases demonstrate the potential of the scheme.
01 Jan 2012
TL;DR: In this paper, a modified HCS detector was proposed to detect ear images under pose variations by using a rotation symmetric, circular detection window, where shape index histograms are extracted at different radii in order to get overlapping subsets within the circle.
Abstract: In identity retrieval from crime scene images, the outer ear (auricle) has ever since been regarded as a valuable characteristic. Because of its unique and permanent shape, the auricle also attracted the attention of researches in the field of biometrics over the last years. Since then, numerous pattern recognition techniques have been applied to ear images but similarly to face recognition, rotation and pose still pose problems to ear recognition systems. One solution for this is 3D ear imaging. the segmentation of the ear, prior to the actual feature extraction step, however, remains an unsolved problem. In 2010 Zhou at al. have proposed a solution for ear detection in 3D images, which incorporates a nave classifier using Shape Index Histogram. Histograms of Categorized Shapes (HCS) is reported to be efficient and accurate, but has difficulties with rotations. In our work, we extend the performance measures provided by Zhou et al. by evaluating the detection rate of the HCS detector under more realistic conditions. This includes performance measures with ear images under pose variations. Secondly, we propose to modify the ear detection approach by Zhou et al. towards making it invariant to rotation by using a rotation symmetric, circular detection window. Shape index histograms are extracted at different radii in order to get overlapping subsets within the circle. The detection performance of the modified HCS detector is evaluated on two different datasets, one of them containing images n various poses.