Keywords

1 Introduction

Fingerprints are the most established form of authentication. Historically, fingerprinting had been near exclusively used in forensics, but over the years, it has grown in popularity where it is now a common method used in various day-to-day services [1, 2]. As such, the need for a better technology to secure a person’s fingerprint identity has also increased [3]. This paper aims to go over the characteristics of a fingerprint that allows for its use in authentication, the technology currently utilized in that endeavor and explore future methods that can be developed for this task.

2 Fingerprint Characteristics

The human fingerprint has two characteristics that makes it ideal for biometric use. The fingerprint is able to generate unique details that enable it for use in identification. The structure of how a fingerprint forms allows for its pattern to remain permanent.

2.1 Identification

Fingerprints are unique due to the shapes and patterns that are formed by ridges. Called minutiae [1], the location of these patterns within the surface of the fingerprints are used to identify a person. It is unlikely for another person to have the exact same kind of minutiae in the exact same location, and this unlikeliness grows exponentially greater with each minutia added for comparison (Fig. 1).

Fig. 1
figure 1

A fingerprint with multiple kinds of minutiae highlighted [4]

Full size image

In addition, the ridges of a fingerprint often follow a specific overall structure. As seen in Fig. 2, these structures, called types, are useful in categorizing fingerprints. An identification system is able to narrow down the number of images that they need to process in order to identify who the print belongs to.

Fig. 2
figure 2

A showcase of the how various types of fingerprints may appear as [5]

Full size image

2.2 Permanence

The human skin is made up of multiple layers but can be simplified to 3 main layers. The epidermis is the outermost layer and acts as a waterproof barrier. The dermis is the middle layer that is mainly composed of hair follicles, sweat glands, and tough connective tissue. The innermost layer is the hypodermis, which is composed of fat and connective tissue.

The fingerprint is formed in the dermis, specifically in the papillary layer that is right beneath the epidermis. This layer anchors to the epidermis using a “double row of peg like protuberances” [6] known as papillae, which forms the layout of the fingerprint. The ridges appear on the epidermis layer are determined by the position of the papillae in the epidermis layer.

This arrangement is what allows fingerprint to remain permanent throughout a person’s life. The papillae act as a “blueprint” that exists under the regenerative layer of the skin. Thus, whenever the epidermal layer restores itself, it will grow following the layout set by the papillae. As long as the papillae remain undamaged, a person’s fingerprint will repair itself to its original design.

3 Authentication System

Fingerprint authentication systems function on three fundamental stages: data acquisition, feature extraction, matching [7]. Data acquisition focuses on the method used by the system in order to extract fingerprint data. Feature extraction is the step where the system identifies unique features of the fingerprint data and stores it within the database. Matching describes the step where a system is required to compare a read fingerprint to those within its database in order to authenticate a person’s identity.

3.1 Data Acquisition

Currently, there are three main types of fingerprint authentication system, defined by their data acquisition methods, optical, capacitive, and ultrasonic fingerprinting. This section will go over their method of acquiring data.

Optical Fingerprinting. Is one of the oldest, cheapest and most common forms of biometric identification [8]. The scanner functions by creating a digital photo of the fingerprint (Fig. 3).

Fig. 3
figure 3

Model of how an optical scanner functions [8]

Full size image

When a fingerprint is placed on the scanning surface, a light source is beamed towards print. An image is generated based on the difference in reflected light levels between ridges and valleys of a fingerprint. The reflected light travels to a light sensor that converts the image into a digital signal. The system performs a few checks to ensure that the generated image is at sufficient brightness before processing the image to be compared to a database of registered fingerprints.

Optical scanners are considered the easiest biometric systems to fool. Its method of simply detecting the changes in reflected light levels means that it can easily be spoofed by simply printing the fingerprint pattern onto a material, such as transparency paper and having the ink produce the ridges and valleys of the fingerprint [8].

Capacitive Fingerprinting. As its name suggest, uses capacitors to “read”s the electrical charges created by the ridges and air pockets of the fingerprint. It is currently the most popular form of fingerprint-based authentication in smartphones [1] (Fig. 4).

Fig. 4
figure 4

Model of how a capacitive scanner functions [8]

Full size image

A capacitive scanner scanning surface is made up of an array of cells that are composed of conductive plates covered in a protective layer [9]. When the ridge of a fingerprint comes in contact with the surface, it acts as another conductive plate, which in turn charges that cell’s conductor plate. The air pockets formed form the valleys however have minimal effect on the charge held by cell. An op-amp integrator is used to detect the changes and the output is recorded by an analog-to-digital converter.

Capacitive scanners, while more secure than optical scanners, have also fallen victim to being spoofed. Simple, 3d molds layered with conductive materials such as gold [10] has been used to successfully bypass most conventional capacitive sensors and complex molds, such as those casted in polydimethylsiloxane, are able to replicate fingerprints at the nanoscale [11].

Ultrasonic Fingerprinting. Is the newest entry in commercial fingerprint detection. They function by generating an ultrasonic pulse onto a finger that is in contact with the scanner. Depending on the details on the fingerprint, some of the pulses are absorbed while others are reflected back to a sensor that is able to detect mechanical stress. The varying intensity of the returned ultrasonic pulse throughout the sensor’s surface creates a detailed 3D reproduction of the scanned fingerprint [9].

Ultrasonic scanners are currently considered of the most secure methods of fingerprint identification, but it has also been spoofed. A researcher was able to create a 3D printed, highly detailed model of a fingerprint to bypass a commercial ultrasonic sensor [12].

3.2 Feature Extraction

After acquiring data through one of the various methods mentioned, the system processes the image to help detect specific features on the fingerprint such as the overall pattern of the ridges or the various minutiae on its surface. The first layer of feature extraction revolves around whether the image remains gray-scale or goes through binarization (Fig. 5).

Fig. 5
figure 5

Classification of minutiae extraction technique [4]

Full size image

Binarization is a process where the gray-scale fingerprint scan is converted to 1’s and 0’s for their respective dark and light areas. This generates a much clearer image of the scan.

Ridge Extraction is a secondary process that can be applied to images that has gone through binarization. The binary image is filtered through a morphological filter that causes the ridges to be thinned [13] to one-pixel thickness. Techniques that utilizes binarization are as follows.

Chaincode processing is a method that traces a ridge along its boundary in a counter-clockwise direction. Ridge endings are detected when the trace makes a significant left turn, while bifurcations are detected when the trace makes a right turn.

Run representation based is a method that performs a vertical and horizontal scan on to the image while applying run-length encoding to the pixels passed. Base on the adjacency of the runs, minutiae are able to be identified [14].

Ridge flow and local pixel analysis process the image through a 3 × 3 mask that calculates the average of each pixel. Pixels that average less than 0.25 denotes a ridge ending while those greater than 0.75 denotes a bifurcation.

Crossing-number based processes the image through a 3 × 3 window that detects the neighboring pixel. Each neighboring pixel adds to the “crossing number” which is a value used to determine the properties of the ridge at that pixel [14] (Fig. 6).

Fig. 6
figure 6

A breakdown of the crossing number method. Each “pattern” within the 3 × 3 window helps the system identify the minutiae [15]

Full size image

Morphology based processes the image through multiple filters that detect a specific shape. An output only appears if the scanned image is able to fully complete the desired shape.

Gray-scale images do have their own advantages over binarization. First of all, it’s able to function with low quality images. Second, binarization and ridge extraction is a time-consuming process, hence gray-scale based processing will often time perform much faster than their processed image counterpart [14]. Techniques based on gray-scale are as follows.

Ridge line following based process functions by following the ridge flow lines set by the fingerprint type. By simply following the lines, the system can detect minutiae it passes through.

Fuzzy based technique uses fuzzy logic model to process the varying gray levels in order to detect minutiae.

3.3 Matching

In order to authenticate a user, the scanned image will be processed by a matching algorithm. This section will briefly describe some of the algorithms used in fingerprint authentication.

Direct matching is the most basic algorithm used for authentication. It functions simply by comparing the pixels of the input image against a template image stored within the database. This method however, requires the system to align the scanned image with the template. This uses a lot of computational power and requires samples to be very large to ensure that the image is aligned properly [16].

Minutiae based matching authentication is performed by detecting minutiae and classifying it based on its surrounding neighbors. The algorithm detects the location, orientation, type and quality of the minutiae [17] and then compares it to ones detected in the print from the database.

Euclidian distance authentication involves identifying 2 minutiae and calculating the distance and angle they are from one another. This process may be repeated for multiple “minutiae pairs”. The system then compares the Euclidian distances of the print and those within its database to find a match [18].

4 System Security

Fingerprint authentication increased use in protecting private information has made the security of the system crucial. A compromised fingerprint has much bigger impact to a person than getting their password stolen since you can’t just change the former as you could the latter.

Threats to the authentication system can be categorized into passive and active attacks. Passive attacks refer to methods that steal information from an authentication system, while active attacks are those that attempt to thwart the authentication service [19, 20]. An example of a passive attack would be a backdoor program that sends a copy of the scanned fingerprint to the attacker, while an example of an active attack would be using the scanned fingerprint to fool a biometric system (Fig. 7).

Fig. 7
figure 7

Flowchart of fingerprint authentication system for both enrollment and authentication function and the various points of attack that can occur [21]

Full size image

5 Future Technology

New methods of fingerprint base authentication are necessary. Currently, there is a research on utilizing thickness of the skin within a fingerprint as a method of biometric identification. While there is no commercial method of thickness-based fingerprinting, there are currently multiple non-invasive methods to measure the thickness of the skin layer. This section aims to describe how these methods work and their potential application into fingerprinting authentication.

Pyroelectric Sensor are passive electronic component that is sensitive to infrared radiation. Changes in temperature causes a change in charge within the pyroelectric crystals within the component, which in turn will generate an electrical signal.

Similar to how a capacitive sensor works, a pyroelectric sensor is able to generate an image based on the location where the ridges of a fingerprint come in contact with the sensor’s surface. The heat generated by the ridges, will be higher than the air pockets within its valleys, which results in differences in the electrical signals generated at those location.

Pyroelectric sensor is potentially a viable method of fingerprinting authentication. The image it generates can easily be implemented into already established extraction and matching processes. A recent research [22] have managed to design a transparent thermal sensor, which increases the ability for its use in commercial smartphones. As seen in Fig. 8, the sensor is able to generate an accurate image of the fingerprint pattern.

Fig. 8
figure 8

Captured fingerprint pattern using pyroelectric sensor [22]

Full size image

Laser Scanning Microscopy (LSM) is an imaging technique that is able to capture multiple micrographs at different depths. It functions by transmitting a filtered, colored light through a pinhole onto the surface of the subject. The subject material is infused with fluorescent chemical that will react with the filtered light. The fluorescent chemical will emit its own filtered light, which will then pass through a dichroic to reach the camera. An image is generated by moving the subject along the x–y axis. The pinhole ensures that the filtered light only activates within a set depth, thus, only an image of that layer. By moving the subject along the z-axis, images for different depths can also be obtained [23, 24].

LSM is deemed a high-resolution technique with a resolution of less than 1 µm, it does have a major drawback. In order for accurate detection to occur, a fluorescent agent must be properly injected into the subject. This results in an increased cost as the fluorescent agent must be resupplied for continuous use.

While it would be extremely difficult for someone to spoof the detection system, as they would have to make a multi-layered model that’s able to replicate each layer exactly, the increased cost and requirement for proper fluorescent agent injection makes it not very viable for authentication use.

Optical Coherence Tomography (OCT) is another non-invasive imaging technique that functions similarly to echo-location. The OCT directs low-coherence light into a tissue. As the light travels throughout the tissue, some of it will scatter when it comes in contact with certain features and materials within the tissue. The OCT then detects the scattered light and utilizes interferometry to filter out photons that have scattered multiple times. This allows for the OCT to create a complete 3D image of the tissue (Fig. 9).

Fig. 9
figure 9

Cross-section of a fingerprint generated through OCT [23]

Full size image

OCT is highly suggested for use in fingerprint authentication, with multiple research [25,26,27] have suggested methods of implementation. OCT’s ability of producing an image of the dermis and epidermis layer of the fingerprint makes it extremely difficult to spoof. Furthermore, the ability for OCT to model individual layers of the skin [28] allows fingerprint thickness to be used as an additional metric for identification.

Near Infrared Spectroscopy (NIRS) is an imaging technique similar to OCT. It functions by beaming near infrared into a tissue sample. The NIRS light will be absorbed differently by different parts of the tissue. A detector is used to measure the change in the reflected light. Information on the thickness of the tissue can be determined by the change in transmittance of the infrared light [29,30,31].

Like OCT, NIRS is able to produce high information scans and is very easy to use. While not as popular as OCT methods of determining fingerprint biometrics, there have been successful research showing its viability to function in authentication [32, 33]. However, in both methods, further testing into its consistency and reaction to various factors, such as skin color, before it can be proven truly viable for commercial use.

6 Conclusion

Fingerprints are essential in the use of biometric based authentication. Its physiology makes it a convenient and ideal tool for that purpose. The number of methodology and techniques used in fingerprint identification is vast, but not without weakness. Thus, alternative methods, such as fingerprint thickness and thermal based detection would be a welcome addition to further improving security.

Of the methods proposed in thickness-based fingerprinting, Optical Coherence Tomography and Near Infrared Spectroscopy are better candidates for further research for authentication application. Their ability to model high detail images of the layers under the epidermis greatly increases the difficulty of being spoofed.