BIOMETRIC - A SURVEY

1.Introduction

The word ‘biometric’ refers to a few unique characteristics of a

person’s physiology or behavior which do not usually change with

time. Examples of such physiological characteristics include

fingerprint, iris, palmprint, face, etc. Examples of behavioral

biometric includes hand-written signature, voice, gait and typing

style on a keyboard.

Biometric can play a major role to verify or to identify an individual.

Verification is the process to confirm the identity of a claimant. In

this process, one or more biometric features of the claimant are

validated against the known biometric profile of the individual.

Therefore, the verification process requires a one-to-one match. In

the case of identifying an individual, the biometric identity of the

unknown individual is matched with the biometric of several others

in an existing database. Hence, identification involves a one-to-many

comparison. Uniqueness of biometric characteristics stops an

imposter against making a false verification and identification

attempts.

The following example can be considered to explain the concept of

the verification process. To avail an Aadhaar-enabled service, the

user types her Aadhaar number to specify her identity. Then, the

system compares the biometric of the user with that of the enrolled

user. Here, one-to-one matching takes place. In this case, an

acceptable similarity between the captured biometric of the

claimant and the biometric of the enrolled user establishes the

genuineness of the claim. Otherwise, the claimant is considered as

an impostor and her claim is rejected. Hence, the verification process

attempts to establish the following: “you are who you say you are”.

On the contrary, in an identification process, an individual claims

that she is one of the registered members as per the record. As a

part of this process, the individual’s biometric is matched with that 

of every member in the record to identify her as one with whom the highest similarity score has been found. But, if the highest similarity score is less than a threshold, then it can be concluded that there is no similarity between the input and the registered members. This establishes the claimant as an impostor. Therefore, an identification process requires 1-to-N comparisons. N is the number of registered members in the above discussion. The identification process establishes an individual as “someone who is already enrolled”. This article revisits various biometric traits, the steps involved in recognition, multimodal biometric systems and the security issues in biometric recognition systems. The organization of the rest of the article is as follows the steps in a biometric recognition system have been depicted in Section 2, various categories of biometric traits are presented in Section 3 and Section 4 discusses about multimodal systems. Fusion in multimodal biometric system can take place at various levels as discussed in Section 5. Section 6 discusses the security concerns supposed to be addressed by a biometric recognition system. Amidst all these theoretical discussions and practical challenges, banks across the globe have embraced biometrics as a factor of authentication. Section 7 provides a glimpse of such adoptions of biometric by banks and the article concludes with Section 8. 2. Steps in Biometric System Registration (or enrollment) and recognition are two phases of a biometric based recognition system. The block diagram is presented in Figure 1. The registration phase includes pre-processing, regionof-interest detection and feature extraction steps. The extracted features are then stored in the database. The recognition phase includes pre-processing, region-of-interest detection, feature extraction, matching and decision making steps. Matching module compares the extracted features with the stored features in the database for either identification or verification task.

2.1. Pre-processing The pre-processing step is primarily used to improve the acquired image (or signal) in order to obtain an accurate extraction of region of interest and the biometric features. Generally, rescaling, perexample mean subtraction and feature standardisation are commonly used as pre-processing for several of the biometric recognition systems. The approach of biometric modality-specific pre-processing is abundant in the literature. Few examples of these are presented below. An experimental study of several illumination pre-processing methods for face recognition is reported in [Han et al, 2013]. These  methods have been divided into three main categories as – graylevel transformation, gradient or edge extraction and reflectance

field estimation. [Jahanbin et al, 2011] has used following four steps

as pre-processing for a face recognition system: gamma correction,

Difference of Gaussian (DoG) filtering, masking and equalization of

variation. The axis-symmetric nature of a face is considered to

generate an approximately symmetrical face image in [Xu et al,

2016] for face recognition. This increases the accuracy of face

recognition methods.

Filtering, equally-spacing, location, size and time normalization are

key pre-processing steps for an online signature verification in

[López-García et al, 2014]. To avoid the acquisition device

dependency, the acquired data is also normalized in a fixed range in

an online signature verification system [Tolosana et al, 2015].

2.2. Finding Region of Interest

Locating the Region of Interest (ROI) for a biometric trait is essential

precursor for feature extraction step. This step identifies the main or

interesting portion of the image (or signal) from where the biometric

traits are extracted. For example, several studies on palmprint

recognition consider the size of the palm to determine the ROI.

Similarly, iris localization is integral to iris recognition [Lili and Mei,

2005].

The methods to extract ROI certainly depends on the modality of the

biometric system. The techniques to identify the region of interest

can be grouped into three major divisions, namely:

(i) Bottom-Up Feature Based Approach: This approach does not

assume any apriori information about the region of interest. Hence,

these approaches are purely driven by detection of key points in a

purely bottom-up approach. For example, face localization can be

carried out using the Scale Invariant Feature Transform (SIFT). A

scale invariant region detector and a descriptor based on the gradient distribution in the detected regions play major role in this approach. (ii) Top-Down Knowledge Based Approach: This approach is influenced by additional relevant information about the physiological characteristics. For example, [Jones and Viola, 2006] has considered individual’s motion and appearance in determining the region of interest. (iii) Appearance Based Approach: This approach considers the inherent physiological appearance of a biometric trait. As an example, it can be studied how region of interest of a palm is extracted in [Saliha et al, 2014]. A key point localization is developed to spot the crucial points of a palm. It helps in the proper alignment of the hand image. This approach is further based on a projection of the X-axis and the projection of the upper and lower edge. Hence, it extracts the horizontal limits of the hand contour. In [Belahcene et al, 2014], a 3D face recognition is proposed by finding regions of interest in a face, which include mouth, nose, pair of eyes, etc. 2.3. Feature Extraction In the feature extraction step, the properties or inherent patterns in a biometric trait is extracted from the input (and possibly preprocessed image/signal). Thus, the derived properties or patterns are a better representation of the unique elements of an individual’s biometric trait. Definitely, the type of biometric decides the feature extraction step. The following paragraph highlights few examples in support of this dependency. Two different feature extraction approaches are present for handwritten signatures as they aim to capture the static or the dynamic features. Geometrical features of the signature are considered for the static approach. Dynamic features of hand-written signature includes the speed and the acceleration of the pen movement, penup and pen-down times, etc. Ear curves are extracted for an earrecognition system in [Ghoualmi et al, 2015]. The face recognition system in [Kumar and Kanhangad, 2015] has used the techniques like wavelet transform, spatial differentiation and twin pose testing scheme for feature extraction from faces. According to [Ukpai et al, 2015], principal texture pattern and dual tree complex wavelet transform produce iris-specific features from an iris image. The next section on various biometric traits will lead to a better understanding of this through narration of different biometric traits. 2.4. Matching and Decision In this step, the extracted features are compared with the enrolled features to obtain a matching score. The subsequent decision making step either accepts or rejects an individual using this matching score. 3. Types of Biometrics 3.1 Fingerprint Uniqueness and consistency in performance have established the fingerprint as the most widely used biometric trait. Usage of fingerprint can be traced back to previous centuries. Ease in acquisition, availability of 10 different fingers and its acceptance for law enforcement and immigration purposes have established it as a very popular form of biometric. A fingerprint is obtained from the friction ridges of the finger. High and peaking part of the skin causes the dark lines in the fingerprint as it is shown in Figure 2. White spaces in between dark lines are due to the shallow parts of the skin, which are also called the valleys. The ridges and furrows (as appearing in Figure 2) enable us to firmly hold objects. Their presence causes a friction, which is needed to grab any object. But uniqueness of fingerprint is not due to these ridges and furrows. Uniqueness is achieved due to minutiae points. The minutiae points are defined as the points where the ridges end, split and join, or appear as a simple dot. The patterns of placement of these minutiae points lead to uniqueness. The minutiae consists of bifurcations, ridge dots, ridge endings and enclosures. The minutiae points are further broken down into sub minutiae such as pores, crossovers and deltas to ensure further uniqueness. Tiny depressions within the ridge are called the pores in a fingerprint. An ‘X’ pattern in the ridge is called crossover. A triangle-shaped pattern in the ridge is called delta. The widespread adoption of fingerprint for biometric recognition is due to several factors like its reasonably good accuracy, ease of use and the small amount of memory space to store biometric template. With the emergence of mobile based applications, the above strengths of fingerprint biometric have led to the use of it for mobile authentication. But the performance of fingerprint recognition drops due to scaly or dirty skin of finger and changes with age. The combination of minutiae points of two different fingers of an individual enables privacy protection in [Li and Kot, 2013].

3.2 Iris Iris is considered to be another reliable biometric trait. The iris is the muscle in the eye which controls the pupil size in order to regulate the amount of light rays entering the eye. The iris can be identified as an annular region in between the sclera (white portion of the eye) and the pupil (Figure 3). The pattern of iris of an individual is also unique [Jain et al, 2004]. A set of twins also possess distinguishing iris patterns. The speed and accuracy of iris recognition have also caused widespread adoption of iris biometric. Registration of iris takes relatively longer time as it needs several iris images. A test template is generated upon scanning of an individual’s iris. Subsequently, the produced template is matched with the existing templates which were produced at the time of registration. Zero crossing representation of the one-dimensional wavelet transform has been proposed in [Radu et al, 2012] to encode the texture in iris. In [Sun et al, 2005], an integration of Local Feature Based Classifier (LFC) and an iris blob matcher increases the accuracy of iris recognition. The noisy images are detected by the iris blob matcher. Therefore, it helps in the situations where the LFC does not guarantee the performance. In [Dong et al, 2011], a set of training images is used to learn classspecific weight map for iris matching technique. Experiments have revealed the effectiveness of the technique. Moreover, a preprocessing method has been suggested to enhance the performance of iris biometric on mobile phones which usually are constrained by computing power. The system has demonstrated its capability to decide whether the person is dead or alive. 3.3 Face The structure of human face is characterized by peaks and valleys at different altitudes and features which are present at different specific latitudes and longitudes as demonstrated in Figure 4. This distinguishes one individual from another. Figure 4: Extraction of Key Features from Human Face Earlier attempts of face recognition used simple geometric models. But sophisticated mathematical representation of features has led to better models of face recognition [Jain et al, 2004]. Combination of certain features with AdaBoost leads to a face and eye detection method in [Parris et al, 2011]. Results are encouraging enough to adopt face based authentication in mobile phones. The face recognition method by [Lai et al, 2014] is assisted with motion sensors. Apple’s iDevice has a face recognition system to lock and unlock it [Gao et al, 2014]. According to [Srinivasan and Balamurugan, 2014], Pictet and Banquiers (one of the leading banks in Switzerland) has deployed an efficient 3D face recognition system for providing access to its staff within the bank’s environment. A graph based model for face recognition has been proposed in [Cao et al, 2012]. Based on 3D features, a face recognition system has been proposed to improve the performance of the system. For detecting facial features, the active appearance model has been used in [Drosou et al, 2012]. A support vector machine based face recognition system has been proposed in [Hayat et al, 2012]. In this technique, an elastic graph matching has been utilized to locate the feature points of the facial image. A face recognition system fails when the face is partly covered. In this case of occlusion, the important characteristics of the face cannot be captured. 3.4 Ear The appearance and shape of the human ear is also found to be unique. It changes little during an individual’s lifetime. Three main steps of an ear biometric system are – (a) imaging of the ear, (b) image segmentation, and (c) recognition. A camera is obviously used for image acquisition. Segmentation is carried out to isolate the ear from the background in the image. A convex curved boundary is identified to locate the ear as in [Maity and Abdel-Mottaleb, 2015]. But experiments have revealed a high false positive due to occlusion. Recognition is performed by comparing the ear biometric traits with stored templates in the database. Local surface patch representation at 3D space leads to a 3D ear recognition system in [Abate et al, 2006]. The ear biometric is captured by Nippon Electric Company (NEC) as the vibration of sound as influenced by the shape of an individual’s ear. It is claimed to be unique for every person. According to this system, an earphone with a built-in microphone captures the sounds as they vibrate within the ear.

3.5 Hand Geometry Measurements of a human hand are used to recognize an individual in the case of hand geometry as biometric trait [Jain et al, 2004]. Shape and size of the palm along with shape, width and length of each finger are considered as important measurements in this context. Edge detectors like Sobel or Canny operators can be used to detect palm lines. Ease of use of this biometric trait leads to wide acceptance of this biometric even in mobile devices [Chen et al, 2007]. But lack of uniqueness of this trait is a major drawback. Hence, its usage is confined only to one-to-one matching. For example, this can be used for access control, where the concern is about an individual’s attempt to gain access through someone else’s access card or personal identification number. The individual’s physical presence is ensured through the presentation of her hand to the hand reader. Though, it can be combined with other biometric traits [Javidnia et al, 2016]. 3.6 Palm Vein Vein pattern in the palm is considered to be another unique trait to recognize an individual. The presence of blood vessels underneath the skin causes this pattern. It is less susceptible to external distortion. Forgery is also difficult for this biometric trait. Moreover, the vein pattern is said to remain static during the lifetime of an individual. The acquisition device throws an infra-red beam on the palm as it is put on the sensor. The veins in the palm are identified as black lines in the captured image. They are matched with an existing vein pattern to recognize an individual [Tome and Marcel, 2015]. Haemoglobin, which is a key ingredient of blood, absorbs the near infra-red (NIR) light. Hence, [Sugandhi et al, 2014] has suggested usage of near infrared (NIR) light to acquire the vein image of fingers. As a result, the vein pattern in fingers appears as shadows. Inspired by fingerprint recognition, palm vein recognition system in [Vaid and Mishra, 2015] also extracts vein minutiae.

3.7 Palmprint A comparatively new biometric trait has been discovered in terms of the palmprint (Figure 5). Reliable and unique characteristics of palmprint justify its high usability. Similar to fingerprint, the palmprint also has unique features, namely, principal lines, minutiae features, delta points, wrinkles, and ridges. Additionally, a wider surface area of the palm (as against the surface area being captured for the fingerprint) leads to more number of unique traits. Hence, palmprint biometric is believed to mature quickly as a reliable recognition system. Figure 5: Palmprint But deformation of images due to challenges of acquisition pulls down the accuracy of a palmprint recognition system. A contact based acquisition device which can pose constraints on acquisition environment is used to solve this problem. Research is still required to tackle the issues arising out of positioning, rotating, and stretching the palm. Moreover, the bigger size of the acquisition device does not allow its usage over mobile phones. Contact based acquisition may also be considered unhygienic. Hence, contactless palmprint acquisition has also been introduced in [Wu and Zhao, 2015]. The users need not touch the acquisition device.3.8 Retina Each individual possesses unique retina vasculature. Replication of it is not easy. The acquisition environment demands an individual to focus her eye on a scanner. Therefore, the system may cause some medical complications like hypertension. This is one reason why this biometric system has not received a full acceptance by the public. 3.9 Radio Biometric Certain physical characteristics (such as height and mass), the condition of the skin, the volume of total body water, and nature of other biological tissues influence the wireless propagation around the human body. Radio biometrics is defined as the identity information as specified by the human-affected wireless signal under alterations and attenuations. The variability of these physical characteristics and biological features among different individuals ensures that two humans are less likely to demonstrate the identical radio biometric. As the chance of two persons having exactly same physical and biological characteristics is very little, the multi-path profiles of the electromagnetic waves after interference from human body vary for each individual. Consequently, human radio biometric, which records how the wireless signal interacts with a human body, are altered according to individuals’ biological and physical characteristics and can be viewed as unique among different individuals. Radio biometric captures the response of radio waves from the entire body including the face of an individual. Hence, it shows more uniqueness than a face. The human identification system in [Xu et al, 2017] uses the entire profile of physical characteristic of an individual. 3.10 Signature Signature defines the way in which one individual writes a specific word (mostly her name or a symbol). It is one of the oldest forms of  biometric and has gained acceptance for several application scenarios. With the progress of technology, two different kinds of signature biometric emerged. Offline signature considers geometrical features of the signature biometric. Online signature provides cue about dynamic features of hand-written signature. These dynamic features include the speed and the acceleration of the pen movement, pen-up and pen-down times, etc. 3.11Gait The posture and the way a person walks maintain uniqueness about the person. It is non-invasive and hard to conceal. It can be easily captured even at public places at a low resolution. Unlike other biometric systems, the individual is not required to pay any attention while her gait is being captured. As the gait can be captured from a distance, it becomes very useful for security. It requires detection of the subject, silhouette extraction, extraction of features, selection of features and classification. A lot of research has been carried out in gait recognition system. Body part (specially, feet and head) trajectories were introduced for extracting gait features. A gait energy image will lead to a set of view intensive features [Liu and Sarkar, 2006]. Recognition is carried out by a match of similar patterns in the gait energy image. Silhouette quality quantification has a key role in the method in [Han and Bhanu, 2006]. A one-dimensional foreground sum signal modeling is used in [Han and Bhanu, 2006] to analyse the silhouettes. Segmenting the human into components and subsequent integration of the results to derive a common distance metric has lead to an improved performance for gait recognition [Vera-Rodriguez et al, 2013]. Using a population based generic walking model, a gait recognition system attempts to solve the challenges which the system encounters due to surface time, movement speed, and carrying condition [Liu and Sarkar, 2005].

The concept of point cloud registration has been proposed in [Lee et al, 2009] in order to avoid the problem of occlusion during gait recognition. Moreover, face and gait characteristics have been extracted using principal component analysis from a side face image and gait energy image, respectively. A set of synthetic features have been attained by integration of these features and application of multiple discriminant analysis. Performance improvement is achieved in comparison to individual biometric features [Tan et al, 2006]. 3.12Voice The voice recognition system extracts several characteristics of voice to identify an individual. Enrollment phase of voice biometric records the voice sample of an individual, extracts a template from it, and uses it for verification of the individual at later phase. Apple’s Siri is a question-answering system based on a voice recognition technology. Mel-frequency cepstral coefficients (MFCC) and support vector machine is used to recognize an individual speaker. One of the drawbacks of this biometric is that a prerecorded voice can easily be played back by an imposter for unauthorized identification. Moreover, a few specific kind of illness (e.g., catching cold) affects the voice and thus, causes hurdle for the voice biometric. 3.13 Key Stroke It is believed that there is a pattern about how a person types on a keyboard. This behavioral biometric, which is referred as key stroke dynamics, shows traits to identify or verify a person. But it is not considered to be as unique as several other biometric traits. 4. Multimodal Biometric System A unimodal biometric system identifies or verifies an individual based on a single biometric trait. Reliability and accuracy of unimodal biometric systems have improved over time. But they always do not demonstrate desired performance in real world applications because of lack of accuracy in the presence of noisy data, non-universal nature of some biometric characteristics, and spoofing. The problems associated with unimodal biometric systems are discussed below. 4.1 Noisy Data Lack of maintenance of sensors may introduce noise within biometric data. For example, typical presence of dirt is common in a fingerprint sensor. It generates a noisy fingerprint. Inability to present the original voice generates a noisy data too. Moreover, iris and face image may not appear as clear without an accurate focus of the camera. 4.2 Non-universality In a universal biometric system, every individual must be capable of producing a biometric trait for recognition. But biometric traits are not always universal. An estimation reveals that about 2% of a population may not be able to produce a good quality fingerprint. Disability of individuals may cause problem for a smooth registration process. Successful enrollment is not possible for such individuals. 4.3 Lack of Individuality Sometimes similar traits are extracted from a biometric system. For example, faces may appear quite similar for father and son, and even more for identical twins. As a consequence of lack of uniqueness, the false acceptance rate increases. 4.4 Susceptibility to Circumvention Sometimes biometric traits are spoofed by an impostor. It is established how fake fingers can be generated by using fingerprints. These can be used to illicitly gain access to a biometric system.

Because of these problems, the error rates are, at times, high for unimodal biometric systems. Hence, they are not always acceptable for security applications. Multimodal biometric systems are conceived to tackle the above mentioned issues. In a multimodal biometric system, multiple biometric features are considered for recognizing an individual. In general, the usage of multiple biometric features contributes in an improved biometric recognition system. For example, a typical error can be caused by worn fingerprints. Presence of other biometric modalities may save the system from failure in the case of multimodal biometric. Thus, multimodal biometric system has less failure to enroll rate. It is considered to be main advantage of multimodal biometric. Multimodal biometric system can be of three types based on how information is fused from various sources of information: (a) fusion of multiple representations of single biometric, (b) fusion of multiple classifiers of single biometric, and (c) fusion of multiple biometrics. Good recognition rate is achieved in a multimodal biometric system involving multiple evidences of a single biometric through fusion of multiple representations or multiple classifiers. But, to a true sense of multimodal biometric system, the use of multiple biometric traits is beneficial than usage of multiple forms of a single biometric in the terms of performance issues, including resistance to low quality samples, lack of individuality, user acceptance, etc. A detailed review of multimodal biometric system can be found in [Oloyede and Hancke, 2016]. Multimodal biometric systems are of three types: 1) multiphysiological, 2) multi-behavioral, and 3) hybrid multimodal systems. In multi-physiological category, only physiological characteristics (for example, fingerprint, retina, face, etc.) are fused. As an example, a multimodal biometric system in [Chang et al, 2003] combines face and ear biometrics. Most of the initial researches in multimodal biometrics belong to this category. Over the past few years, the rapid developments in human-machine interface have triggered an evolution in behavioral biometric recognition. Hence, the field of behavior based multimodal biometric system has drawn attention of many researchers. A multi-behavioral biometric system in [Fridman et al, 2013] considers inputs from mouse, keyboard, writing sample, and history of web browsing. In [Bailey et al, 2014], another multibehavioral biometric system considers inputs from graphical user interface interactions alongside mouse and keyboard inputs. Moreover, a hybrid multimodal biometric system combines physiological and behavioral features. Notable works on hybrid multimodal biometric include fusion of face, audio and speech using multiple classifiers by [Fox et al, 2007], fusion of face and gait by [Tan et al, 2006], and signature, face, and ear biometric fusion at the score level by [Monwar and Gavrilova, 2009]. Another good hybrid multimodal biometric system in [Paul et al, 2014] combines signature, face and ear biometric alongside social network analysis. It has been shown here that inputs from social network analysis further strengthen the biometric recognition system. Contextual information (such as spatiotemporal information, appearance, and background) has a key role alongside soft biometrics (for example, height, weight, facial marks, and ethnicity) in identifying a person [Park and Jain, 2010]. In [Bharadwaj et al, 2014], face biometric and social contextual information have shown a significant improvement over performance in a challenging environment. It is to be noted that neither of extraction of appropriate contextual information or acquisition of soft biometric are easy tasks. These tasks may even require image processing. Moreover, social behavioral information is a common contributor in the normal recognition process in the human brain. [Sultana et al, 2017] administers a reinforcing stimulus in the form of social behavioral information to the matching decisions of traditional face and ear based biometric recognition system.

5. Levels of Fusion in Multimodal Biometric System A detailed classification of various fusion techniques for multimodal biometric can be found in [Dinca and Hancke, 2017]. Fusion in multimodal biometric can occur at various levels – such as sensor, feature, matching score, rank and decision level fusion. Each of these is explained in this section. 5.1 Sensor Level Fusion This fusion strategy directly mixes raw data from various sensors (Figure 6) – for example, from the iris and fingerprint sensors. Raw information is captured at several sensors to fuse at the very first level to generate raw fused information. Sensor level fusion strategies can be put into following three groups: (i) single sensor multiple instances, (ii) intra-class multiple sensors, and (iii) inter-class multiple sensors. In the case of single sensor multiple instances, a single sensor captures multiple instances of the same biometric trait. For example, a fingerprint sensor may capture multiple images of the same finger to reduce the effect of noise. Simple or weighted averaging, and mosaic construction are some of the common fusion methods in this case [Yang et al, 2005]. Multiple sensors are used to acquire multiple instances of the same biometric trait in the intraclass multiple sensors category [Yang et al, 2005]. For example, a 3D face image is obtained by using multiple face images taken from various cameras. In the case of inter-class multiple sensors, two or more different biometric traits are used together. For example, images of palmprint and palm vein can be fused together for biometric recognition. Mosaicing is a nice application of sensor level fusion. Several researchers [Ratha et al, 1998; Jain and Ross, 2002; Ross et al, 2005] have proposed fingerprint mosaicing. It provides a good recognition accuracy as it combines multiple images of the same fingerprint. Therefore, it can handle the difficulty in recognition due to data quality. The fingerprint mosaicing technique uses a modified Iterative Closest Point (ICP) [Jain and Ross, 2002] algorithm to generate 2D or 3D surfaces by considering the inputs from multiple instances. In [Fatehpuria et al, 2006], a touchless fingerprint system is developed using a 3D touchless setting with multiple cameras and structured light illumination (SLI) to generate 2D fingerprint images and 3D fingerprint shape. This kind of set up is expensive due to deployment of multiple cameras. Alternatively, usage of a single camera and two mirrors are suggested in [Choi et al, 2010]. Two mirrors have been used to obtain finger side views. Sensor level fusion is generally applied for the same trait. There are also instances of applying sensor level fusion for different traits. Few of these are mentioned here. Face and palmprint images are combined in [Jing et al, 2007]. Pixel level fusion is preceded by Gabor transform of the images. Infrared images of palmprint and palm vein are fused in [Wang et al, 2008]. At first, image, registration is carried out on these images. Subsequently, a pixel level fusion takes place. 5.2 Feature Level Fusion Features extracted from several biometric traits are integrated into a single vector. According to this fusion strategy, biometric sensor signals (from camera or microphone) are preprocessed and feature vectors are derived from them independently. Then a composite feature vector is generated by combining these individual feature vectors (Figure 7). The feature level fusion exhibits a better performance than score level and decision level fusion techniques as feature level fusion techniques directly deals with the unique biometric features. Normalization and selection of features are two important processes in feature level fusion. Min-max technique and media scheming based normalization is carried out to change the scale and location of feature values. Scale invariant feature transform is also carried out from the normalized images. Figure 7: Feature Level Fusion Dimensionality reduction through appropriate feature selection also enhances the accuracy of the techniques. Sequential forward selection, sequential backward selection, and partition about medoids are standard feature selection techniques. Particle Swarm Optimization (PSO) is applied on the feature vector for dimensionality reduction. The multimodal biometric techniques in [Raghavendra et al, 2009; 2011] uses this concept while combining face and palmprint features.

Incompatibility of the feature sets among different biometric traits and non-linearity of the joint feature set of different biometric traits poses challenges for feature level fusion. A feature vector can be generated using weighted average of multiple feature vectors if those vectors correspond to same biometric. For example, this becomes possible if all of these vectors are obtained from fingerprint images of an individual. If these vectors correspond to different biometrics, then they are concatenated to obtain a single vector. Another example of feature level fusion can be found in [Kim et al, 2011]. Simultaneous use of time-of-flight (ToF) depth camera and near infrared (NIR) camera acquires face and hand vein images in a touchless acquisition set up. Several multimodal system also combines face and ear, as ear is considered to be one of the most unchangeable feature of the human traits. Unlike face, human ear is not generally affected by age. PCA based feature extraction and a sparse representation method for feature level fusion is proposed in [Huang et al, 2013]. Experimental results reveal that this technique performs better than their unimodal components. Experiments also show that the performance is similar to that of the unimodal systems even if one of the modality is corrupted. Local 3D features (L3DF) are generated from ear and frontal face images in [Islam et al, 2013]. Feature level fusion is applied in these cases. A matrix interleaved concatenation based new approach is presented in [Ahmad et al, 2016] for face and palmprint biometrics. Discrete Cosine Transform (DCT) is used here to extract the features. Then, these features are concatenated in an interleaved matrix which estimates the parameters of the feature concatenation and exhibits their statistical distribution. A fingerprint and iris based multimodal biometric recognition technique has been proposed in [Gawande et al, 2013]. Minutiae and wavelet features are extracted from fingerprint images. Haar wavelet and block sum techniques produce features from iris images. A feature level fusion of these four feature vectors exhibit better performance than a unimodal fingerprint or iris biometric. A feature level fusion of fingerprint and palm biometric traits has been proposed in [Mohi-ud-Din et al, 2011]. Another example of feature level fusion can be found for hand geometry recognition in the contactless multi-sensor system in [Svoboda et al, 2015] using an Intel RealSense 3D camera. This technique carries out foreground segmentation of the acquired hand image to determine the hand silhouette and contour. Then, the fingertips and the valleys are located alongside determination of the wrist line from the identified contour. Subsequently, two features vectors have been formed as follows: (i) comprising finger length, width, and wrist valley distances and (ii) finger widths as computed using a traversal of the overall hand surface and median axis to surface distances. A finger and finger-vein based system has been proposed in [Yang and Zhang, 2012]. Gabor features are extracted and the feature fusion strategy is based on a Supervised Local-Preserving Canonical Correlation Analysis (SLPCCAM). In [Yan et al, 2015], feature level fusion has also been used for a contactless multi-sample palm vein recognition technique. Automated access control systems in buildings and other secure premises are based on the capability of identifying an individual from a distance. Because of its importance in securing important establishments, it is emerging as an area of interest to the research community. Improper lighting condition or not-so-high resolution surveillance cameras pose a constraint for recognizing an individual based on her face. Use of multimodal biometric involving face and gait exhibits better performance [Ben et al, 2012; Huang et al, 2012]. Unlike traditional methods of face and gait multimodal biometric recognition, [Xing et al, 2015] fuses the features without normalization using coupled projections. A Robust Linear Programming (RLP) method [Miao et al, 2014] for multi-biometric recognition exhibits good results in noisy environments in spite of using less training data. It uses uncertain constraints and concatenates heterogeneous features from different biometric traits. Each biometric modality has been assigned a weight to specify its degree of contribution in the fusion. More weight is a greater relevance of the corresponding biometric trait. In the multimodal biometric recognition system by [Chang et al, 2003], features are extracted from face and ear using Principle Component Analysis (PCA). Subsequently, fusion takes place at the feature level. A method of liveliness detection to prevent spoofing attack in [Chetty and Wagner, 2005] also uses a feature level fusion. In a recent development, sparse-based feature-fusion [Huang et al, 2015] of physiological traits has drawn sufficient interest of researchers due to robust performance. 5.3 Matching Score Level Fusion In the case of matching score level fusion, matching scores are separately obtained for each biometric trait and subsequently fused to arrive at an overall matching score. The block diagram is presented in Figure 8. Matching score level fusion is also referred as measurement level fusion. There exists three different approaches for matching score based fusion – density based, classification based, and transformation based. The density based scheme is on the basis of distribution of scores and its application to popular models like Naive Bayesian and Gaussian Mixture Model [Murakami and Takahashi, 2015]. In the classification based approach, the matching scores of individual matching modules are concatenated to obtain a single feature vector. The decision to either accept or reject an individual is based on the classification of this feature vector. According to the transformation based approach, the scores of individual matching module are, at first, transformed (normalized) into a pre-decided range. This transformation changes the position and scale parameters of the matching score distribution so that these normalized scores can be combined to obtain a single scalar score [Murakami and Takahashi, 2015]. These normalization techniques to handle the dissimilarities in matching score also draw attention of researchers. In another noted development of recent time, an order-preserving score fusion method has been proposed in [Liang et al, 2016].

5.4 Rank Level Fusion Ranking the potential matches between the query template and the templates in the database generates an ordered list of all templates in the database. The first choice is the match. These ranked are obtained for every biometric trait. In a multimodal biometric recognition system, these rank orders are fused to generate a final ranking of each template in the database. Unlike score level fusion, normalization is not required for rank level fusion. Rank level fusion is applied in [Kumar and Shekhar, 2011] for combining various methods for palmprint identification. A Nonlinear Weighted Ranks (NWR) method aggregates the ranks as obtained from individual matching modules. Rank level fusion may not always perform well in noisy conditions having low quality data. Though it has been applied on low quality fingerprints [Abaza and Ross, 2009]. It applies a derivation of the Borda count method, which involves image quality. This approach has a similarity with logical regression. But unlike logical regression, this approach does not need a training phase. Image quality is considered in this approach instead of weights. As ranks are assigned to only few of the stored templates with possible match, the rank level fusion may create a challenge for large databases. Ranks do not cover every template in the database. In this context, a Markov chain based method has been proposed in [Monwar and Gavrilova, 2011] for rank level fusion. Markov chain is used to represent a stochastic series of events, where the present or the preceding states determine the next state. A graph is used to formally model the Markov chain. A vertex in the graph represents a state or an event. An edge in the graph denotes the transition from one state to another state. At first, ranks are generated for each biometric trait. If the matching module creates partial ranking (for example, the first three ranking results), elements are inserted randomly to complete the list. The state transition probabilities are computed and the stationary distribution of the Markov chain is obtained. The templates in the database are ranked based on a decreasing order of the scores of the stationary distribution starting from the highest score. This fusion strategy is applied on a multimodal biometric recognition system involving iris, ear, and face. [Monwar and Gavrilova, 2009] proposes another method to solving this discussed problem of rank level fusion. The multimodal biometric recognition system has three matchers for each of signature, face and ear. However, fusion is carried out between the identities put out by at least two matchers. 5.5 Decision Level Fusion In the case of a decision level fusion, each individual matcher, at first, takes its own decision. Subsequently, the fusion of various biometric modalities takes place by combining the decisions of these individual matchers. Hence, each biometric trait is independently pre-classified and the final classification is based on the fusion of the outputs of the various modalities (Figure 9). The simplest forms of decision level fusion uses logical operations such as ‘AND’ or ‘OR’. Some advanced fusion strategies at this level also use behavior knowledge space, the Dempster-Shafer theory of evidence, and Bayesian decision fusion. Figure 9: Decision Level Fusion

When every individual decision module supplies positive outcome, then the ‘AND’ rule positively recognizes a query template. Otherwise, the ‘AND’ rule rejects the query. Hence, ‘AND’ rule is generally reliable with extremely low false acceptance rate (FAR). But false rejection rate (FRR) is higher than that of individual trait. On the contrary, the ‘OR’ rule provides positive output about a query template when at least one decision module gives positive response about it. As a result, FRR is extremely low and FAR is higher than individual trait. In [Tao and Veldhuis, 2009], an optimized threshold method has been proposed using the ‘AND’ and ‘OR’ rule. The thresholds of the classifiers are optimized during the training phase. Majority voting is another common approach for decision fusion. If the majority of the individual traits decide positively about the query template, then final decision is positive. Majority voting method gives equal importance to each individual decision modules. Otherwise, a weighted majority voting can be applied. In this method, higher weights are assigned to decision modules which perform better. A multi-algorithm decision level fusion is used in [Prabhakar and Jain, 2002] for fingerprints. This method considers four distinct fingerprint matching algorithms. These are based on Hough transform, string distance, 2D dynamic programming, and texture. This method selects appropriate classifiers prior to applying decision fusion. A threshold for each individual classifier influences the outcome of a decision level fusion. Here, threshold specifies a minimum score to decide whether the sample is genuine or an impostor. If the matching score of the sample is higher than the threshold, then the sample is considered as genuine. On the contrary, if the matching score of the sample is less than the threshold, then it belongs to an imposter. The classifiers are assumed to be independent of one another in some biometric systems. However, other works have assumed the dependency among the classifiers. A verification system has been introduced in [Veeramachaneni et al, 2008] based on two fusion strategies for correlated threshold classifiers. Between these two strategies, Likelihood Ratio Test (LRT) still depends on the threshold of each individual classifier. The Particle Swarm Optimisation (PSO) based decision strategy is considered effective in comparison. Even the PSO strategy performs better than some of the score level fusion methods. A real time sensor management using PSO is suggested in [Veeramachaneni et al, 2005] for a multimodal biometric management. This method performs a real time search of the optimal sensor configuration and optimal decision rule. A similar concept is proposed in [Kumar et al, 2010] which uses Ant Colony Optimization (ACO) technique, for a multimodal biometric system involving palmprint and hand vein. [Kumar and Kumar, 2015] extends this experiment on multiple multimodal databases involving palmprint and iris, fingerprint and face, and face and speech. Another decision level fusion for multimodal biometric recognition is proposed in [Paul et al, 2014]. In this multimodal system, signature, face and ear biometric features are combined with social network analysis. The Fisher image feature extraction, which is a combination of PCA and Linear Discriminant Analysis (LDA). 5.6 Hybrid Fusion Model A hybrid fusion model which uses both pixel level fusion and score level fusion demonstrates good performance in [Kusuma and Chua, 2011]. This multi-sample face recognition (both in 2D and 3D) in [Kusuma and Chua, 2011] recombines images using principal component analysis (PCA). Two recombined images are fused using a pixel level fusion scheme. Additionally, score level fusion is also applied to produce good result. 6. Security and Privacy Issues in Biometric There are several security and privacy concerns associated with usage of biometric. These are listed below: 

• Biometric is not secret. Biometric data can be captured by a third party very easily. At times, even the original user may not be aware of the spying of her biometric data. For example, voice recording during a telephonic conversation or by a rogue mobile app can disclose the user’s voice biometric. Similarly, a video recording in the guise of surveillance cameras captures the user’s face or gait biometric • Biometric cannot be changed or revoked. Unlike password or pin, it is impossible to issue a new biometric trait, e.g., fingerprint. It is permanent. One-time compromise makes it unusable. Moreover, the user can never be dissociated with the compromised data • Biometric can be used to track an individual forever. There exists eight different ways to attack a biometric system [Ratha et al, 2001]. These are demonstrated in Figure 10 with numbers which are explained below: 1. Fake biometric: Fake biometric is presented to the sensor with an intention of fooling the system. There have been several successful demonstrations of this kind of attack. 2. Resubmitting stored signals: The biometric signal can be pre-recorded and can be presented to the system at later times 3. Overriding feature extraction process: The feature extraction module is compromised. The actual feature set is replaced with the desired one by an attacker. 4. Tampering with biometric representation: The templates representing the actual biometric trait are replaced with a desired one by the attacker 5. Corrupting the matcher: The matching module is compromised and the attacker generates a matching score as desired by herself 

Compromising stored templates: An attacker may illicitly gain access to the stored templates. She can steal those templates to spoof the users’ identities 7. Communication interception: The information being passed between the matcher and the database can be altered if an attacker intercepts the communication between these two modules 8. Overriding the final decision: An attacker may override the decision being taken by the matcher.

All the above points indicate how a biometric recognition system can be compromised. But most of the fraudulent attempts take place in the context of faking a biometric data or tampering with the stored template. Fake biometrics can be recreated from the stored templates. Even it can be acquired directly from the sensor without the knowledge of the user. Moreover, compromising the database may lead to editing or deleting of the templates. These two topics are discussed here at length. 6.1 Faking Biometric Data Numerous researches reveal how biometrics can be faked. Following are the three easy steps to spoof fingerprint data: (i) The residual fingerprint of a user can be obtained from a mobile phone or any other surface; (ii) Lifted fingerprint impression can be used to create a dummy finger; (iii) The dummy finger can be put on the fingerprint sensor to claim the identity. Demonstrations of this kind of attack against the fingerprint sensors of popular smart phone brands are publicly available [web1]. Biometric data is often captured in Internet of Things (IoT) devices with or without user’s knowledge. Hence, these devices represent the most danger to biometric. Biometric identity can be spoofed using the captured data. Recently, 5.6 million fingerprint records have been stolen from Office of Personnel Management (OPM) of US military. Such a large scale breach reveals devastating consequences of poor data security practices. Moreover, it strengthens the concern over storage of biometric data, as the OPM unit of US military is considered to enforce stringent security practices while storing biometric data in comparison to several private companies. Apple’s Siri, Google Now, and Microsoft Cortana record every uttering of a user and sends the data back to the servers of their organisations through Internet [Sparkes, 2015]. Samsung TVs automatically record conversations of the users to use these for automatic speech recognition [Matyszczyk, 2015]. Behavioral biometric is captured by many of the wearable devices. Several forms of biometric data can also be captured through a smartphone. Hence, smartphones pose a risk to privacy. Though misuse of biometric data by big corporations is debatable, numerous third party applications installed in smartphones may appear as security and privacy risk. These third party applications often ask for more permission in the device than what is actually needed for them to complete their tasks. Permissions for accessing the camera and the microphone of a smartphone are the most misused ones [Felt et al, 2011]. These permissions enable the application to capture face, retina, and voice samples of the user. Moreover, many applications request for root permission to access every device sensor such as fingerprint, gait, heart monitoring and key logging for getting behavioral information about keystrokes and screen shots. These are even very handy in spying for user credentials. It is debatable whether these third party applications use the data with malicious intensions. But their business models, in several cases, allow them to sell the users’ data to advertisers. With the increasing demand of users’ data in underground market, it may be more lucrative to sell users’ data than to make money through in-app advertising. As per a 2012 report [Labs, 2012] from Zcaler’s labs, over 40% of mobile applications communicate data to third parties. During installation of a new application, majority of the smartphone users do not check the permission requests. Even many of them may not be aware of the implications. Additionally, the entire permission system can be bypassed using root exploits [Zhou and Jiang, 2012]. Moreover, storage of data at third party servers poses a risk. Past cases show that there have been breaches even in military servers. Then, breaching the security of a small smartphone application vendor may not be challenging for the attackers. A comprehensive review of biometric authentication in a smartphone is presented by [Meng et al, 2015]. Twelve types of biometric authentication can be carried out in a smartphone. Among those, following six are physiological: fingerprint, iris, retina, face, hand and palmprint. Remaining six types are behavioral in nature. These are signature, gait, voice, keystroke dynamics, touch dynamics, and behavioral profiling. A survey of successful attacks on smartphones is presented in this report [Meng et al, 2015]. Another successful attack to guess passwords using touch devices is also reported in [Zhang et al, 2012]. The proposals for securing smartphone authentication schemes, and authentication in general, are use of multimodal biometrics, check for liveness, combining with other authentication techniques (dual factor authentication) and use cancelable biometrics to store the templates. These proposals are useless if the user installs malware or a free application including a root kit that bypasses the permission system and captures biometric data from the user as he uses the smartphone.

Several problems exist with smartphone permission across various operating systems. Experts have also suggested improvements in this regard. But all of these efforts in security may fail at a single point, i.e., the user. This is because the user, sometimes, installs rogue applications without proper checking of the source or vendor. There exist numerous fake applications which pretend to be popular ones. Even the users, at several times, do not read the permissions while installing the applications. Hence, all good security practices fail. Certain permissions to an application provide the application an access to several device features and implicitly to data – biometric or otherwise. Even if the application owner is not misusing the data, there may be a breach by a malicious attacker to steal the data for illicit gains. An artificial finger can spoof a fingerprint on many fingerprint sensors. Several researchers have demonstrated it in time and again [Cappelli et al, 2007; Galbally et al, 2010; Espinoza et al, 2011]. Research also suggests how this spoofing attack on fingerprints can be prevented [Marcialis et al, 2010]. Similarly, iris biometric is also vulnerable to spoofing through fake iris scans. Several techniques [Wei et al, 2008; Rigas and Komogortsev, 2015] suggests how to detect a fake iris. Another biometric very susceptible to spoofing attacks is face authentication by using pictures. There are a lot of techniques proposed to detect this issue [Maata et al, 2011; Komulainen et al, 2013; Pereira et al, 2014]. Hands geometry can also be spoofed by creating fake hands. [Chen et al, 2005] who proposed a practical model using plaster to create fake hands. The authors demonstrate that the fake hands can be created without the user knowledge from hand templates stored into the database. Other soft biometrics can be easily spoofed – voice can be easily recorded or spoofed artificially [Alegre et al, 2012], gait can be spoofed using a video camera from a distance to capture the user motion [Gafurov et al, 2007].

6.2 Template Security Until one decade ago, it was believed that a stored template cannot recreate the original biometric data. Several researchers proved this wrong [Ross et al, 2007; Jain et al, 2008]. Encryption cannot be used to prevent a template compromise, as it is not possible to carry out recognition in the encrypted domain [Jain et al, 2008]. Tamper resistant storage in a smart card seems feasible for a single template for verification. Otherwise, it cannot be applied to large biometric databases. Solutions exist in the form of private templates [Davida et al, 1998] and cancelable biometrics [Ratha et al, 2007]. Still, several biometric recognition systems have not adopted these solutions to secure the templates in the database. The concept of cancelable biometric to tackle the above stated problem was first proposed in [Ratha et al, 2007]. Several other template protection schemes have been developed subsequently. These schemes can be grouped into following two categories – cancelable transformations and biometric cryptosystems. The characteristics of a template protection scheme are mentioned here: • Diversity: The template of a biometric of same individual has to be distinct in different databases. It will prevent an attacker from gaining access to multiple systems through a compromise in one database • Revocability: In the case of a compromise in an individual’s biometric template, it will be possible to issue a new template to her from the same biometric data • Security: It will not be possible to recreate the original biometric data from a template. It is a one-way transformation • Performance: There should not be any impact on the performance of the biometric system in terms of false acceptance rate (FAR) and false rejection rate (FRR).

[Jain et al, 2008] describes the advantages and disadvantages of each template protection type. There is extensive literature on cancelable uni-biometric schemes and cryptosystems thoroughly surveyed by [Rathgeb and Uhl, 2011]. Even a subsequent chapter in this Staff Series discusses the issue of cancelable biometric at depth. 6.3 Other Types of Attacks on Biometric Systems Similar to a brute force approach of password guessing, a brute force attack with a large number of input fingerprints can be used. The difficulty of such an attack is that the search space for guessing the fingerprint is prohibitively large. But for fingerprint authentication on several mobile devices, only a part of the full fingerprint is utilized. This provides the attacker with a much smaller search space. On the contrary, a dictionary attack tries only those possibilities which are deemed most likely to succeed. Although dictionary attacks have been extensively studied and analyzed for traditional password-based authentication systems, they have not been systematically considered by the research community in the context of fingerprint verification. To perform a guessing attack with fingerprints, the question arises as to whether there are some fingerprints that are more likely to match a target than the others? It has been observed in the previous literature, that different users have different performance characteristics based on their fingerprint. [Yager and Dunstone, 2010] has introduced a menagerie consisting of dove (users with high genuine scores and low imposter scores), chameleons (high genuine scores and high imposter scores, thus are easy to match with everyone, including themselves), phantom (hard to match with most of the users), and worm (hard to match with themselves but easy to match with others). [Yager and Dunstone, 2010] have identified the existence of chameleons in datasets of full fingerprints.

A metric to estimate the strength of a biometric recognition system against impersonation attacks, namely Wolf Attack Probability (WAP), is proposed in [Une et al, 2007]. In this context, a wolf indicates an input sample which wrongly matches with multiple biometric templates. [Roy et al, 2017] shows how a master print can be located or generated. Such a master print can be used to match with multiple biometric templates. In case of partial fingerprints, the probability of detecting a master print and the attack accuracy increase. This finding reveals the risks of usage of partial fingerprints for authentication. In [Nagar et al, 2012], an evidential value analysis is carried out for latent fingerprints. It has been observed that less number of minutiae points or a small surface area of the latent fingerprint leads to a low evidential value of the fingerprint. Hence, probability of matching error increases in these cases. Based on this finding, an estimate in [Roy et al, 2017] shows that the probability of finding masterprints that incorrectly match with a large number of templates is high for partial fingerprints. 7. Usage of Biometric in Banks and Financial Institutions Amidst all these practical challenges and concerns over security and privacy issues in biometric, banks and financial institutions have taken significant steps in embracing it. Banks in India have also embraced biometric as one of the authentication factor. Here is a small illustrative list of initiatives by banks in India: 1. DCB Bank: Fingerprint based cash withdrawal is possible from the ATMs of DCB Bank. These ATMs connects with Aadhaar database to authenticate a customer using her fingerprint as enrolled with Aadhaar. 2. Federal Bank: A zero balance selfie account opening is possible with Federal Bank using its banking app. For an instantaneous account opening, a user scans her Aadhaar and PAN cards and then, clicks her selfie photo. 3. HDFC Bank: As part of financial inclusion initiatives by HDFC Bank, the bank has introduced fingerprint verification using a hand-held device or a micro-ATM. Fingerprint based verification with the Aadhaar database enables the bank for instant KYC (know your customer) check for its users. 4. ICICI Bank: Voice recognition enables the customers of ICICI Bank to interact smoothly with the bank’s call center. During such an interaction, authentication credentials are not being asked to the customers, as their voice can authenticate themselves. 5. State Bank of India: A fingerprint based authentication is carried out in order to provide the bank’s employees an access to the core banking system. There are several use cases of biometric in banking and financial service industry around the globe too. Fingerprint is commonly being used by mobile banking applications to authenticate a user for last few years. For example, Bank of America, Chase and PNC are some of the institutions which have adopted fingerprint based user authentication for their mobile applications. Master Card has launched a ‘selfie-pay’ to authenticate online purchases through either face or fingerprint recognition. Citi has registered its customers’ voice samples. USAA, which serves to members of the military and their families in United States, has rolled out three different biometrics – fingerprint, face and voice recognition – for customer authentication. Pictet and Banquiers (one of the leading banks in Switzerland) has deployed an efficient 3D face recognition system for providing access to its staff within the bank’s environment. 8. Conclusion Biometric recognition is gaining popularity for identification and verification of individuals through their specific physiological and behavioral traits. In certain scenarios, its importance is perceived in the form of a second factor of authentication, in addition to knowledge based or possession based authentication requiring security for transactions. It enables the government as well as private and public businesses to reduce identity theft and related crimes. In this article, the strengths and weaknesses of several biometric recognition systems have been discussed through a comprehensive review of the developments in this field. Furthermore, both unimodal and multimodal biometric systems have been discussed. Various types of fusion strategies have also been explained in the context of multimodal biometric systems. This article also discusses various security and privacy concerns which are associated with usage of biometrics. A compilation of the progress in this field helps the readers get an overall grasp. 

References [web1] “Chaos Computer Club Breaks Apple TouchID”, http://www.ccc.de /en/updates/2013/ccc-breaksapple-touchid, January 2016 A. F. Abate, M. Nappi, D. Riccio, and S. Ricciardi, “Ear Recognition by Means of a Rotation Invariant Descriptor,” in Proc. 18th International Conference on Pattern Recognition (ICPR), pp. 437-440, August 2006 A. Abaza and A. Ross, “Quality Based Rank-Level Fusion in Multibiometric Systems,” in Proc. IEEE 3rd International Conference on Biometrics, Theory, Applications and Systems (BTAS), pp. 1-6, September 2009 M. I. Ahmad, W. L. Woo, and S. Dlay, “Non-Stationary Feature Fusion of Face and Palmprint Multimodal Biometrics,” Neurocomputing, Vol. 177, pp. 49-61, February 2016 F. Alegre, R. Vipperla, N. Evans, and B. Fauve, “On The Vulnerability of Automatic Speaker Recognition to Spoofing Attacks with Artificial Signals,” in Proc. 20th European Signal Processing Conference (EUSIPCO), pp. 36-40, August 2012 K. O. Bailey, J. S. Okolica, and G. L. Peterson, “User Identification and Authentication Using Multi-Modal Behavioral Biometrics,” Computer Security, Vol. 43, pp. 77–89, June 2014

M. Belahcene, A. Chouchane, and H. Ouamane, “3D Face Recognition in Presence of Expressions by Fusion Regions of Interest,” in Proc. 22nd Signal Processing and Communications Applications Conference (SIU), pp. 2269- 2274, April 2014 X. Y. Ben, M. Y. Jiang, Y. J. Wu, and W. X. Meng, “Gait Feature Coupling for Low-Resolution Face Recognition,” Electronics Letters, Vol. 48, No. 9, pp. 488-489, April 2012 S. Bharadwaj, M. Vatsa, and R. Singh, “Aiding Face Recognition with Social Context Association Rule Based Re-Ranking,” in Proc. IJCB, Clearwater, pp. 1–8, FL, USA, 2014 X. Cao, W. Shen, L. G. Yu, Y. L. Wang, J. Y. Yang, and Z. W. Zhang, “Illumination Invariant Extraction for Face Recognition Using Neighboring Wavelet Coefficients,” Pattern Recognition, Vol. 45, No. 4, pp. 1299-1305, 2012 R. Cappelli, D. Maio, A. Lumini, and D. Maltoni, “Fingerprint Image Reconstruction from Standard Templates,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, No. 9, pp. 1489-1503, September 2007 K. Chang, K. W. Bowyer, S. Sarkar, and B. Victor, “Comparison and Combination of Ear and Face Images in Appearance-Based Biometrics,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 25, No. 9, pp. 1160–1165, September 2003 W.-S. Chen, Y.-S. Chiang, and Y.-H. Chiu, “Biometric Verification by Fusing Hand Geometry and Palmprint,’ in Proc. 3rd International Conference on Intelligent Information Hiding and Multimedia Signal Processing (IIHMSP), pp. 403-406, November 2007 H. Chen, H. Valizadegan, C. Jackson, S. Soltysiak, and A. K. Jain, “Fake Hands: Spoofing Hand Geometry Systems,” Biometric Consortium, Washington DC, USA, 2005 G. Chetty and M. Wagner, “Liveness Detection Using Cross-Modal Correlations in Face-Voice Person Authentication,” in Proc. 9th European Conference on Speech Communication and Technology (Eurospeech), Lisbon, Portugal, pp. 2181–2184, 2005

H. Choi, K. Choi, and J. Kim, “Mosaicing Touchless and Mirror-Reflected Fingerprint Images,” IEEE Transactions on Information Forensics and Security, Vol. 5, No. 1, pp. 52-61, March 2010 M. Cornett, “Can Liveness Detection Defeat the M-Commerce Hackers?”, Biometric Technology Today, Vol. 2015, No. 10, pp. 9-11, October 2015 G. I. Davida, Y. Frankel, and B. Matt, “On Enabling Secure Applications through Off-Line Biometric Identification,” in IEEE Symposium on Security and Privacy, pp. 148-157, May 1998 L. M. Dinca and G. P. Hancke, “The Fall of One, the Rise of Many: A Survey on Multi-Biometric Fusion Methods”, IEEE Access, Vol. 5, pp. 6247-6289, April 2017 W. Dong, Z. Sun, and T. Tan, “Iris Matching Based on Personalized Weight Map,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 33, No. 9, pp. 1744-1757, September 2011 A. Drosou, N. Porfyriou, and D. Tzovaras, “Enhancing 3D face Recognition Using Soft Biometrics,” in Proc. 3DTV Conference: The True Vision – Capture, Transmission and Display of 3D Video, pp. 1-4, October 2012 M. Espinoza, C. Champod, and P. Margot, “Vulnerabilities of Fingerprint Reader to Fake Fingerprints Attacks,” Forensic Science International, Vol. 204, Nos. 1-3, pp. 41-49, January 2011 A. Fatehpuria, D. L. Lau, and L. G. Hassebrook, “Acquiring a 2D Rolled Equivalent Fingerprint Image from a Non-Contact 3D Finger Scan,” Proc. SPIE, Vol. 6202, pp. 62020C, April 2006 A. P. Felt, E. Chin, S. Hanna, D. Song, and D. Wagner, “Android Permissions Demystified,” in Proc. 18th ACM Computer and Communications Security Conference, New York, USA, pp. 627-638, 2011 N. A. Fox, R. Gross, J. F. Cohn, and R. B. Reilly, “Robust Biometric Person Identification Using Automatic Classifier Fusion of Speech, Mouth, and Face Experts,” IEEE Transactions on Multimedia, Vol. 9, No. 4, pp. 701–714, June 2007 A. Fridman et al., “Decision Fusion for Multimodal Active Authentication,” IT Professional, Vol. 15, No. 4, pp. 29–33, 2013

J. Galbally et al., “An Evaluation of Direct Attacks using Fake Fingers Generated from ISO Templates,” Pattern Recognition Letters, Vol. 31, No. 8, pp. 725-732, June 2010 Z. Gao, D. Li, C. Xiong, J. Hou, and H. Bo, “Face Recognition with Contiguous Occlusion Based on Image Segmentation,” in Proc. International Conference on Audio, Language and Image Processing (ICALIP), pp. 156-159, July 2014 U. Gawande, M. Zaveri, and A. Kapur, “Bimodal Biometric System: Feature Level Fusion of Iris and Fingerprint,” Biometric Technology Today, Vol. 2013, No. 2, pp. 7-8, 2013 L. Ghoualmi, A. Draa, and S. Chikhi, “An Efficient Feature Selection Scheme Based on Genetic Algorithm for Ear Biometrics Authentication,” in Proc. 12th International Symposium on Programming and Systems (ISPS), pp. 1-5, April 2015 M. Gomez-Barrero, J. Galbally, and J. Fierrez, “Efficient Software Attack to Multimodal Biometric Systems And Its Application To Face And Iris Fusion,” Pattern Recognition Letters, Vol. 36, pp. 243-253, January 2014 J. Han and B. Bhanu, “Individual Recognition Using Gait Energy Image,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 28, No. 2, pp. 316-322, February 2006 H. Han, S. Shan, X. Chen, and W. Gao, “A Comparative Study On Illumination Preprocessing in Face Recognition,” Pattern Recognition, Vol. 46, No. 6, pp. 1691-1699, June 2013 M. Hayat, M. Bennamoun, and A. A. El-Sallam, “Fully Automatic Face Recognition from 3D Videos,” in Proc. 21st International Conference on Pattern Recognition (ICPR), pp. 1415-1418, November 2012 Z. Huang, Y. Liu, X. Li, and J. Li, “An Adaptive Bimodal Recognition Framework Using Sparse Coding for Face and Ear,” Pattern Recognition Letters, Vol. 53, pp. 69–76, February 2015 Z. Huang, Y. Liu, C. Li, M. Yang, and L. Chen, “A Robust Face and Ear Based Multimodal Biometric System Using Sparse Representation,” Pattern Recognition, Vol. 46, No. 8, pp. 2156-2168, August 2013 Y. Huang, D. Xu, and F. Nie, “Patch Distribution Compatible Semi-Supervised Dimension Reduction for Face and Human Gait Recognition,” IEEE 

Transactions on Circuits and Systems for Video Technology, Vol. 22, No. 3, pp. 479-488, March 2012 S. M. S. Islam, R. Davies, M. Bennamoun, R. A. Owens, and A. S. Mian, “Multibiometric Human Recognition Using 3D Ear and Face Features,” Pattern Recognition, Vol. 46, No. 3, pp. 613-627, March 2013 S. Jahanbin, H. Choi, and A. C. Bovik, “Passive Multimodal 2-D+3-D Face Recognition using Gabor Features and Landmark Distances,” IEEE Transactions on Information Forensics and Security, Vol. 6, No. 4, pp. 1287- 1304, December 2011 A. K. Jain, S. C. Dass, and K. Nandakumar, “Soft Biometric Traits for Personal Recognition Systems,” in Biometric Authentication, Berlin, Germany: Springer, pp. 731-738, 2004 A. K. Jain, K. Nandakumar, and A. Nagar, “Biometric Template Security,” EURASIP Journal on Advances in Signal Processing, Vol. 2008, pp. 113, April 2008 A. Jain and A. Ross, “Fingerprint Mosaicking,” in Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Vol. 4, pp. IV-4064-IV-4067, May 2002 H. Javidnia, A. Ungureanu, C. Costache, and P. Corcoran, “Palmprint as a Smartphone Biometric,” in Proc. IEEE International Conference on Consumer Electronics (ICCE), pp. 463-466, January 2016 X.-Y. Jing, Y.-F. Yao, D. Zhang, J.-Y. Yang, and M. Li, “Face and Palmprint Pixel Level Fusion and Kernel DCV-RBF Classifier for Small Sample Biometric Recognition,” Pattern Recognition, Vol. 40, No. 11, pp. 3209-3224, November 2007 M. J. Jones and P. Viola, “Method and System for Object Detection in Digital Images,” US Patents 7 099 510, August 29, 2006 T.-C. Kim, K.-M. Kyung, and K. Bae, “New Biometrics-Acquisition Method Using Time-Of-Light Depth Camera,” in Proc. IEEE International Conference on Consumer Electronics (ICCE), pp. 721-722, January 2011 J. Komulainen, A. Hadid, M. Pietikainen, A. Anjos, and S. Marcel, “Complementary Countermeasures for Detecting Scenic Face Spoofing Attacks,” in Proc. International Conference on Biometrics (ICB), pp. 1-7, June 2013

A. Kumar, M. Hanmandlu, H. Sanghvi, and H. M. Gupta, “Decision Level Biometric Fusion Using Ant Colony Optimization,” in Proc. 17th IEEE International Conference on Image Processing (ICIP), pp. 3105-3108, September 2010 T. S. Kumar and V. Kanhangad, “Face Recognition Using Two-Dimensional Tunable-Q Wavelet Transform,” in Proc. International Conference on Digital Image Computing: Techniques and Applications (DICTA), pp. 1-7, 2015 A. Kumar and A. Kumar, “Adaptive Management of Multimodal Biometrics Fusion Using Ant Colony Optimization,” Information Fusion, Vol. 32, pp. 49- 63, November 2015 A. Kumar and S. Shekhar, “Personal Identification Using Multibiometrics Rank-Level Fusion,” IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), Vol. 41, No. 5, pp. 743-752, September 2011 G. P. Kusuma and C.-S. Chua, “PCA-based Image Recombination for Multimodal 2D C 3D Face Recognition,” Image and Vision Computing, Vol. 29, No. 5, pp. 306-316, April 2011 Z. Labs., “10% of Mobile Apps Leak Passwords, 40% Communicate with Third Parties j Cloud Security Solutions j Zscaler” [Online]. Available: https://www.zscaler.com/press/10-mobile-apps-leak-passwords-40- communic%ate-third-parties, 2012 K. Lai, S. Samoil, and S. N. Yanushkevich, “Multi-spectral Facial Biometrics in Access Control,” in Proc. Symposium on Computational Intelligence in Biometrics and Identity Management (CIBIM), pp. 102-109, December 2014 T. K. Lee, M. Belkhatir, and S. Sanei, “On the Compensation for the Effects of Occlusion in Fronto-Normal Gait Signal Processing,” in Proc. 11th IEEE International Symposium on Multimedia (ISM), pp. 165-170, December 2009 S. Li and A. C. Kot, “Fingerprint Combination for Privacy Protection,” IEEE Transactions on Information Forensics and Security, Vol. 8, No. 2, pp. 350- 360, February 2013 Y. Li, K. Xu, Q. Yan, Y. Li, and R. H. Deng, “Understanding OSN-based Facial Disclosure Against Face Authentication Systems,” in Proc. ACM Symposium 

on Information, Computer and Communications Security, New York, USA, pp. 413-424, 2014 Y. Liang, X. Ding, C. Liu, and J.-H. Xue, “Combining Multiple Biometric Traits with an Order-Preserving Score Fusion Algorithm,'' Neurocomputing, Vol. 171, pp. 252-261, January 2016 P. Lili and X. Mei, “The Algorithm of Iris Image Preprocessing,” in Proc. 4th IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), pp. 134-138, October 2005 Z. Liu and S. Sarkar, “Effect of Silhouette Quality On Hard Problems in Gait Recognition,” IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), Vol. 35, No. 2, pp. 170-183, April 2005 S. Maity and M. Abdel-Mottaleb, “3D Ear Segmentation and Classification Through Indexing,” IEEE Transactions on Information Forensics and Security, Vol. 10, No. 2, pp. 423-435, February 2015 G. Marcialis, F. Roli, and A. Tidu, “Analysis of Fingerprint Pores for Vitality Detection,” in Proc. 20th International Conference Pattern Recognition (ICPR), pp. 1289-1292, August 2010 C. Matyszczyk, “Samsung’s Warning: Our Smart TVs Record Your Living Room Chatter” [Online]. Available: http://www.cnet.com/news/samsungswarning-our-smart-tvs-recordyour-liv%ing-room-chatter/, 2015 W. Meng, D. Wong, S. Furnell, and J. Zhou, “Surveying the Development of Biometric User Authentication on Mobile Phones,” IEEE Communications Surveys & Tutorials, Vol. 17, No. 3, pp. 1268-1293, 3rd Quarter, 2015 D. Miao, Z. Sun, and Y. Huang, “Fusion of Multi-Biometrics Based on a New Robust Linear Programming,” in Proc. 22nd International Conference on Pattern Recognition (ICPR), pp. 291-296, August 2014 S.-U.-D. G. Mohi-ud-Din, A. B. Mansoor, H. Masood, and M. Mumtaz, “Personal Identification Using Feature and Score Level Fusion of Palm and Fingerprints,” Signal, Image and Video Processing, Vol. 5, No. 4, pp. 477- 483, August 2011 M. M. Monwar and M. L. Gavrilova, “Multimodal Biometric System Using Rank-Level Fusion Approach,” IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, Vol. 39, No. 4, pp. 867–878, August 2009

M. M. Monwar and M. Gavrilova, “Markov Chain Model for Multimodal Biometric Rank Fusion,” Signal, Image and Video Processing, Vol. 7, No. 1, pp. 137-149, April 2011 T. Murakami and K. Takahashi, “Information-Theoretic Performance Evaluation of Likelihood-Ratio Based Biometric Score Fusion under Modality Selection Attacks,” in Proc. 7th International Conference on Biometrics Theory, Applications and Systems (BTAS), pp. 1-8, September 2015 A. Nagar, H. Choi, and A. K. Jain, “Evidential Value of Automated Latent Fingerprint Comparison: An Empirical Approach,” IEEE Transactions on Information Forensics and Security, Vol. 7, No. 6, pp. 1752–1765, December 2012 M. O. Oloyede and G. P. Hancke, “Unimodal and Multimodal Biometric Sensing Systems: A Review”, IEEE Access, Vol. 4, pp. 7532-7555, September 2016 U. Park and A. K. Jain, “Face Matching and Retrieval Using Soft Biometrics,” IEEE Transactions on Information Forensics and Security, Vol. 5, No. 3, pp. 406–415, September 2010 J. Parris et al., “Face and Eye Detection on Hard Datasets,” in Proc. International Joint Conference on Biometrics (IJCB), pp. 1-10, October 2011 P. P. Paul, M. L. Gavrilova, and R. Alhajj, “Decision Fusion for Multimodal Biometrics Using Social Network Analysis,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, Vol. 44, No. 11, pp. 1522–1533, November 2014 M. Paunwala and S. Patnaik, “Biometric Template Protection with DCT Based Watermarking,” Machine Vision and Applications, Vol. 25, No. 1, pp. 263-275, July 2013 T. D. F. Pereira et al., “Face Liveness Detection Using Dynamic Texture,” EURASIP Journal on Image and Video Processing, Vol. 2014, No. 1, pp. 1-15, January 2014 N. Poh, J. Kittler, C.-H. Chan, and M. Pandit, “Algorithm to Estimate Biometric Performance Change Over Time,” IET Biometrics, Vol. 4, No. 4, pp. 236-245, 2015 S. Prabhakar and A. K. Jain, “Decision-level Fusion in Fingerprint Verification,” Pattern Recognition, Vol. 35, No. 4, pp. 861-874, 2002

P. Radu, K. Sirlantzis, W. Howells, S. Hoque, and F. Deravi, “Image Enhancement Vs Feature Fusion in Colour Iris Recognition,” in Proc. International Conference on Emerging Security Information, Systems and Technologies (EST), pp. 53-57, September 2012 R. Raghavendra, B. Dorizzi, A. Rao, and G. Hemantha, “PSO versus AdaBoost for Feature Selection in Multimodal Biometrics,” in Proc. IEEE 3rd International Conference on Biometrics: Theory, Applications and Systems (BTAS), Sep. 2009, pp. 1-7. R. Raghavendra, B. Dorizzi, A. Rao, and G. H. Kumar, “Designing Efficient Fusion Schemes for Multimodal Biometric Systems Using Face and Palmprint,” Pattern Recognition, Vol. 44, No. 5, pp. 1076-1088, May 2011 N. K. Ratha, S. Chikkerur, J. H. Connell, and R. M. Bolle, “Generating Cancelable Fingerprint Templates,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, No. 4, pp. 561-572, April 2007 N. K. Ratha, J. H. Connell, and R. M. Bolle, “Image Mosaicing for Rolled Fingerprint Construction,” in Proc. 14th International Conference on Pattern Recognition, Vol. 2, pp. 1651-1653, August 1998 N. K. Ratha, J. H. Connell, and R. M. Bolle, “Enhancing Security and Privacy in Biometrics-based Authentication Systems,” IBM Syst. J., Vol. 40, No. 3, pp. 614-634, April 2001 C. Rathgeb and A. Uhl, “A Survey on Biometric Cryptosystems and Cancelable Biometrics,” EURASIP Journal on Information Security, Vol. 2011, No. 1, pp. 1-25, 2011 I. Rigas and O. V. Komogortsev, “Eye Movement-Driven Defense against Iris Print-Attacks,” Pattern Recognition Letters, Vol. 68, pp. 316-326, December 2015 

R. N. Rodrigues, N. Kamat, andV. Govindaraju, “Evaluation of Biometric Spoofing in a Multimodal System,” in Proc. 4th IEEE International Conference on Biometrics Theory, Applications and Systems (BTAS, pp. 1-5), September 2010 A. Ross, S. Dass, and A. Jain, “A Deformable Model for Fingerprint Matching,” Pattern Recognition, Vol. 38, No. 1, pp. 95-103, 2005

A. Ross, J. Shah, and A. K. Jain, “From Template to Image: Reconstructing Fingerprints from Minutiae Points,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 29, No. 4, pp. 544–560, April 2007 

A. Roy, N. Memon, and A. Ross, “Masterprint: Exploring The Vulnerability Of Partial Fingerprint-Based Authentication Systems”, IEEE Transactions on Information Forensics and Security, Vol. 12 ( 9), pp. 2013-2025, September 2017 

A. Saliha, B. Karima, K. Mouloud, D. H. Nabil, and B. Ahmed, “Extraction Method of Region of Interest from Hand Palm: Application with Contactless and Touchable Devices,” in Proc. 10th International Conference on Information Assurance and Security (IAS), pp. 77-82, November 2014 

S. Shah, A. Ross, J. Shah, and S. Crihalmeanu, “Fingerprint Mosaicing Using Thin Plate Splines,” in Proc. Biometric Consortium Conference, 2005, pp. 1-2 H. M. Sim, H. Asmuni, R. Hassan, and R. M. Othman, “Multimodal Biometrics: Weighted Score Level Fusion Based On Nonideal Iris and Face Images,” Expert Systems with Applications, Vol. 41, No. 11, pp. 5390-5404, September 2014 

M. Sparkes, “Why Your Smartphone Records Everything You Say To It”, [Online], Available: http://www.telegraph.co.uk/technology/news/11434754/Why-yoursmartphone%-records-everything-you-say-toit.html, 2015 

A. Srinivasan and V. Balamurugan, “Occlusion Detection and Image Restoration in 3D Face Image,” in Proc. IEEE Region 10 Conference (TENCON), pp. 1-6, October 2014 

N. Sugandhi, M. Mathankumar, and V. Priya, “Real Time Authentication System Using Advanced Finger Vein Recognition Technique,” in Proc. International Conference on Communication and Signal Processing (ICCSP), pp. 1183-1187, April 2014 

M. Sultana, P. P. Paul, M. L. Gavrilova, “Social Behavioural Information Fusion in Multimodal Biometrics”, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2017 

Z. Sun, Y. Wang, T. Tan, and J. Cui, “Improving Iris Recognition Accuracy Via Cascaded Classifiers,” IEEE Transactions on Systems, Man, and Cybernetics Part C: Applications and Reviews, Vol. 35, No. 3, pp. 435-441, August 2005

Y. Sutcu, Q. Li, and N. Memon, “Secure Biometric Templates from Fingerprint-Face Features,” in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1-6, June 2007 

J. Svoboda, M. Bronstein, and M. Drahansky, “Contactless Biometric Hand Geometry Recognition Using a Low-Cost 3D Camera,” in Proc. International Conference on Biometrics (ICB), pp. 452-457, May 2015 

D. Tan, K. Huang, S. Yu, and T. Tan, “Efficient Night Gait Recognition Based On Template Matching,” in Proc. 18th International Conference on Pattern Recognition (ICPR), August 2006 

Q. Tao and R. Veldhuis, “Threshold-Optimized Decision-Level Fusion and Its Application to Biometrics,” Pattern Recognition, Vol. 42, No. 5, pp. 823-836, May 2009 

K. Tewari and R. L. Kalakoti, “Fingerprint Recognition and Feature Extraction Using Transform Domain Techniques,” in Proc International Conference on Advanced Communication Control and Computing Technologies (ICACACT), pp. 1-5, August 2014 

R. Tolosana, R. Vera-Rodriguez, J. Ortega-Garcia, and J. Fierrez, “Preprocessing and Feature Selection for Improved Sensor Interoperability in Online Biometric Signature Verification,” IEEE Access, Vol. 3, pp. 478-489, 2015 

P. Tome and S. Marcel, “On The Vulnerability of Palm Vein Recognition to Spoofing Attacks,” in Proc. International Conference on Biometrics (ICB, pp. 319-325), May 2015 

C. O. Ukpai, S. S. Dlay, and W. L. Woo, “Iris Feature Extraction Using Principally Rotated Complex Wavelet Filters (PR-CWF),” in Proc. International Conference Computer Visual Image Analysis Application (ICCVIA), pp. 1-6, January 2015

 U. Uludag, S. Pankanti, S. Prabhakar, and A. Jain, “Biometric Cryptosystems: Issues and Challenges,” Proc. IEEE, Vol. 92, No. 6, pp. 948-960, June 2004 

M. Une, A. Otsuka, and H. Imai, “Wolf Attack Probability: A New Security Measure in Biometric Authentication Systems,” in Proc. International Conference on Biometrics, pp. 396–406, 2007

S. Vaid and D. Mishra, “Comparative Analysis of Palm-Vein Recognition System Using Basic Transforms,” in Proc. International Conference on Current Trends in Advanced Computing (IACC), pp. 1105-1110, June 2015 

K. Veeramachaneni, L. Osadciw, A. Ross, and N. Srinivas, “Decision Level Fusion Strategies for Correlated Biometric Classifiers,” in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Workshops (CVPRW), pp. 1-6, June 2008 

K. Veeramachaneni, L. A. Osadciw, and P. K. Varshney, “An Adaptive Multimodal Biometric Management Algorithm,” IEEE Transactions on Systems, Man, and Cybernetics, Part C, Vol. 35, No. 3, pp. 344-356, August 2005 

R. Vera-Rodriguez, J. Fierrez, J. S. D. Mason, and J. Orteua-Garcia, “A Novel Approach of Gait Recognition Through Fusion with Footstep Information,” in Proc. International Conference on Biometrics (ICB, pp. 1-6), June 2013 

K. Vishi and S. Y. Yayilgan, “Multimodal Biometric Authentication Using Fingerprint and Iris Recognition in Identity Management,” in Proc. 9th International Conference Intelligent Information Hiding and Multimedia Signal Processing, pp. 334-341, October 2013 

N. Wang, Q. Li, A. A. A. El-Latif, X. Yan, and X. Niu, “A Novel Hybrid Multibiometrics Based On the Fusion of Dual Iris, Visible and Thermal Face Images,” in Proc. International Symposium on Biometrics and Security Technologies (ISBAST), pp. 217-223, July 2013 

J.-G. Wang, W.-Y. Yau, A. Suwandy, and E. Sung, “Person Recognition by Fusing Palmprint and Palm Vein Images Based On ‘Laplacianpalm’ Representation,” Pattern Recognition, Vol. 41, No. 5, pp. 1514-1527, May 2008 

Z. Wei, X. Qiu, Z. Sun, and T. Tan, “Counterfeit Iris Detection Based On Texture Analysis,” in Proc. 19th International Conference on Pattern Recognition (ICPR), pp. 1-4, December 2008 

P. Wild, P. Radu, L. Chen, and J. Ferryman, “Robust Multimodal Face and Fingerprint Fusion in the Presence of Spoofing Attacks,” Pattern Recognition, Vol. 50, pp. 17-25, September 2016

X. Wu and Q. Zhao, “Deformed Palmprint Matching Based On Stable Regions,” IEEE Transactions on Image Processing, Vol. 24, No. 12, pp. 4978- 4989, December 2015 

X. Xing, K. Wang, and Z. Lv, “Fusion of Gait and Facial Features Using Coupled Projections for People Identification at a Distance,” IEEE Transactions on Signal Processing, Vol. 22, No. 12, pp. 2349-2353, December 2015 

Q. Xu, Y. Chen, B. Wang, and K. J. R. Liu, “Radio Biometrics: Human Recognition Through a Wall”, IEEE Transactions on Information Forensics and Security, Vol. 12 (5), pp. 1141-1155, May 2017 

Y. Xu, L. Fei, and D. Zhang, “Combining Left and Right Palmprint Images for More Accurate Personal Identification,” IEEE Transactions on Image Processing, Vol. 24, No. 2, pp. 549-559, February 2015 

Y. Xu, Z. Zhang, G. Lu, and J. Yang, “Approximately Symmetrical Face Images for Image Preprocessing in Face Recognition and Sparse Representation Based Classification,” Pattern Recognition, Vol. 54, pp. 68-82, June 2016 

X. Yan, W. Kang, F. Deng, and Q. Wu, “Palm Vein Recognition Based On Multi-Sampling and Feature-Level Fusion,” Neurocomputing, Vol. 151, pp. 798-807, March 2015 

F. Yang, M. Paindavoine, H. Abdi, and A. Monopoli, “Development of a Fast Panoramic Face Mosaicking and Recognition System,” Optical Engineering, Vol. 44, No. 8, pp. 087005-1_087005-10, 2005 

J. Yang and X. Zhang, “Feature-level Fusion of Fingerprint and Finger-vein for Personal Identification,” Pattern Recognition Letters, Vol. 33, No. 5, pp. 623-628, April 2012 

Y.-F. Yao, X.-Y. Jing, and H.-S. Wong, “Face and Palmprint Feature Level Fusion for Single Sample Biometrics Recognition,” Neurocomputing, Vol. 70, Nos. 7-9, pp. 1582-1586, March 2007 

N. Yager and T. Dunstone, “The Biometric Menagerie,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 32, No. 2, pp. 220–230, February 2010 

Y. Zhang, P. Xia, J. Luo, Z. Ling, B. Liu, and X. Fu, “Fingerprint Attack Against Touch-enabled Devices,” in Proc. 2nd ACM Workshop on Security & Privacy in Smartphones and Mobile Devices, New York, USA, pp. 57-68, 2012

Y. Zhou and X. Jiang, “Dissecting Android Malware: Characterization and Evolution,” in Proc. IEEE Symposium on Security and Privacy (SP), pp. 95- 109, May 2012.