Open access peer-reviewed chapter

A Survey of Machine Learning Techniques for Behavioral-Based Biometric User Authentication

By Nurul Afnan Mahadi, Mohamad Afendee Mohamed, Amirul Ihsan Mohamad, Mokhairi Makhtar, Mohd Fadzil Abdul Kadir and Mustafa Mamat

Submitted: February 11th 2018Reviewed: March 22nd 2018Published: October 31st 2018

DOI: 10.5772/intechopen.76685

Downloaded: 408


Authentication is a way to enable an individual to be uniquely identified usually based on passwords and personal identification number (PIN). The main problems of such authentication techniques are the unwillingness of the users to remember long and challenging combinations of numbers, letters, and symbols that can be lost, forged, stolen, or forgotten. In this paper, we investigate the current advances in the use of behavioral-based biometrics for user authentication. The application of behavioral-based biometric authentication basically contains three major modules, namely, data capture, feature extraction, and classifier. This application is focusing on extracting the behavioral features related to the user and using these features for authentication measure. The objective is to determine the classifier techniques that mostly are used for data analysis during authentication process. From the comparison, we anticipate to discover the gap for improving the performance of behavioral-based biometric authentication. Additionally, we highlight the set of classifier techniques that are best performing for behavioral-based biometric authentication.


  • continuous authentication
  • behavioral biometric
  • machine learning
  • classification
  • clustering

1. Introduction

Over the past decade, the field of computer security has evolved along with the changing nature of technology. Computer security comprises of measures and controls that ensure the goals of information security that are confidentiality, integrity, and availability, defined over hardware, software, firmware, and information being processed, stored, and communicated, are achieved [1]. These goals of information security, also known as CIA triad, is a benchmark model used to evaluate the physical, logical, and perceptual security of information in an organization [2, 3]. The elements of the triad are considered as the three most crucial components of information security. It can have serious effects for an organization if any one of this triad is breachable.

Confidentiality is roughly equivalent to privacy or secrecy which offers prevention of the sensitive information from disclosure by unauthorized individuals or systems [4]. By and large, it is also the one which is attacked most often. Cryptography via encryption algorithms is commonly used to ensure the confidentiality of data in storage or transferred from one computer to another.

Integrity is typically described as the trustworthiness, accuracy, and consistency of data in which the data itself cannot be altered or modified undetectable by unauthorized user [4]. Cryptography plays a major role in ensuring data integrity. This is done by hashing the original data and transmitting the data and the hash to the recipient followed by another hashing on the received data and comparison with the received hash to verify its integrity.

Availability is defined as the security controls required to ensure that the information concerned is readily accessible to the authorized parties when they request it [1]. Denial of service (DoS) attack can be a good example of many threats to this security controls. DoS renders the system to an unavailable state to serving legitimate request by making the server fully utilized the processing power, bandwidth, and memory to handle request mostly mounted by this attack.

Last but not least, authentication is a key point to provide effective information security. Authentication process verifies the identity of a user, process, or device and allows only legal users to use the resources and services in an authorized manner while denying all illegal ones [1].

Nowadays, user authentication is an issue and thus a challenge that becomes more important than ever before [5]. For an online banking system, it is very important to secure the users’ accounts and protect their assets and personal information from malicious hands due to highly sensitiveness of data held inside. There are many existing authentication methods; in general, they are categorized into knowledge-based method, possession-based method, and biometric-based method. For sure, all of the methods have their own uniqueness (strengths and weaknesses); however, the environment determines which authentication approach is best suited.

When talking about the authentication in general, two types of well-known approaches have been proposed in the literature, namely, continuous authentication approach and static authentication approach [6]. Continuous authentication approach which can also be acknowledged as dynamic authentication verifies users repeatedly throughout the entire session [7]. The benefit of this approach is that the system is able to continuously monitor if there is any unauthorized access that occurs.

Meanwhile, static authentication approach collect data from the user and verify their access and privileges in manipulating the data, for example, at the login time [7]. This accessing service will be valid until the user logs out from the session. The combination of username and password is a popular method for static authentication. Nevertheless, there is a drawback for static authentication in which this approach will authenticate the user only at the beginning of each session. The system will remain unnoticeable if there is any change of user in case of attacks [6].

In this paper, we survey the most recent advancement in biometric authentication system. However, our focuses are only on behavioral-based biometric authentication. In order to evaluate the accuracy of behavioral-based biometric authentication [8], there are three common measurements which are false rejection rate (FRR), the percentage of users’ wrongly denied access to a system; false acceptance rate (FAR), the percentage of users wrongly authorized by a system; and equal error rate (EER), the value of the FRR and FAR when a system is tuned to have an equal FAR and FRR. Generally, in order for the authentication system to be more practical, it must have the following features that are accuracy, quick response, and difficult to be forged [9].

The remainder of this article is organized as follows: Section 2 discusses the biometric authentication. The subtopic in this section described the taxonomy of user authentication methods in each category emphasizing on their advantages and disadvantages. The description of behavioral-based biometric authentication system for every paper is discussed in details. Section 3 presented a discussion and future research direction in the development of behavioral-based biometric authentication system. Finally, Section 4 concludes the paper.

2. Authentication

2.1. User authentication methods

The most important key for the authentication process is the uniqueness of security measures, which in general can be categorized into something a user knows (password), something a user has (smart card), or something a user is (biometrics) [10, 11, 12, 13]. Some examples of knowledge-based method, possession-based method, and biometric-based method can be found in Figure 1.

Figure 1.

Taxonomy of user authentication methods.

2.1.1. Knowledge-based method

Knowledge-based technique is commonly used to secure the access for systems [14]. The two famous examples are the pin and password. The password is normally entered at the beginning of any communication or operation which is only allowed if user has the correct one. The benefits for using conventional password are no specialized personnel required, simple, easy to use, and easy to remember. Unfortunately, passwords have many problems in that it is highly vulnerable to brute force attacks, password guessing, and key-loggers. The drawback is that once the password is compromised, an opponent can easily exploit a victim’s account [15].

The marbles gap approach which comprises of password in a form of arbitrary sequence of marbles during authentication process can be found in [16]. The user needs to drag the digits in the right direction into the center of the screen. After that, it immediately reappears on the prior position. In order to leave smudge traces, three graphic-based authentication methods were implemented, which are one grid-based and two randomized graphical approaches.

Another authentication scheme for smartphone was established by using the matrix values of image [17]. This approach requires synchronization in advance between the smartphone and the service server. For the task of authentication, the user must react to the service server by inputting an existing combination of text-based and graphic-based passwords and thus providing better accuracy.

Ref. [18] proposed a location-based authentication approach using smartphone. The static (captured at login session) and continuous (captured during the session) location information were used. The two different locations of APIs were utilized during location verification. The location was verified and compared prior deciding whether the user is valid or not. This system can provoke errors during verification process caused by overlapping in location. Therefore, the security of the system introduced depends on the effectiveness of location verification.

Ref. [19] presented the physical proximity to guarantee security using a modulated illumination of smartphone screens to transmit PIN. The user enters a PIN on smartphone. By using a cheap bespoke receiver unit, the PIN is transmitted via temporary patterns of light on the screen. This approach was the right choice to ensure confidentiality against man-in-the-middle attacks.

The hybrid graphical password approach which is the mixture of recall and recognition-based schemes provided more secure system according to the use of graphical and textual password [20]. During registration phase, the user selects a username and a textual password and then chooses an object as password by drawing. All the information is stored in a database. During the authentication process, the user enters username and textual password and then draws the preselected objects. As expected, this scheme is not intended for users without drawing capability.

Table 1 shows a summary of various existing user authentication schemes that falls under knowledge-based category listed with advantages and disadvantages. Due to these advantages, the area of knowledge-based method for user authentication becomes less unpopular for exploration by the researchers.

[16]Graphical passwordThe marbles authentication methodThis method has no upper restriction for the password spaceThe pattern of key arrangement must be recognized by the user
[17]Graphical passwordMatrix values of imageProvides more accuracy caused by the combination of sensorsPower consumption
[18]LocationLocation-based authentication
  • Used the mobile function

  • Easy to use

Can provoke errors in verification caused by the overlapping in location
[19]PINA modulated illumination of mobile device screens to transmit PINAssures confidentiality against attacksLight sensor works within limited geographic scope
[20]Graphical passwordRecall and recognition-based schemesMore secure caused by the combination of graphical and textual passwordCan provoke login error if the user does not have drawing capability

Table 1.

Summary of knowledge-based method.

2.1.2. Possession-based method

The usages of traditional password have already been indicated as not sufficiently secure and inconvenient as a security measure. The possession-based method was proven to eradicate the risk of an attacker to guess passwords and is predicted to raise the level of security to data. This method makes use of things the user personally possesses such as token, smart card, and QR code.

Any objects or devices that can be used during authentication process are called hardware tokens. They are available in various forms such as a mobile device [21] or an easy-access device (key fobs and smartphones). The smart card reader (NFC-enabled smartphones) approach has been introduced with the combination of PIN and smart card [22]. The PIN is managed as a temporary PIN. The use of temporary PIN reduces the chance for an attacker to distinguish the permanent PIN.

The user authentication using QR code identification approach was implemented in this system [23]. During verification phase, the user makes a request from the server; in return, the server will extract the information about that user. The benefit of this approach is that it is known to be faster than the certificate system.

A summary of possession-based category is shown in Table 2. Possession-based methods are proven to eradicate the risk of an attacker to guess passwords easily from knowledge-based method. Since the token is needed to be present during the authentication process, the drawbacks of physical token are that, from the stolen or lost token, an attacker might gain an authorized access. Thus, the possession-based method for user authentication can still be considered as weak.

[22]PIN + smart cardSmart card reader (NFC-enabled smartphones)The use of a temporary pin will reduce the chance for an attacker to detect the permanent pinPublic terminal or computer is required as an input and output device for smart cards
  • Acoustic token

  • Magnetic token

Sound waves and static magnetic fieldsLess prone to snoopingA sharp drop in the strength of the magnetic field formed can cause complications to the user
[23]QR codeQR code identifying for user authentication
  • Easy to use

  • Low cost

  • Reduces the memorization of human

Table 2.

Summary of possession-based method.

2.1.3. Biometric-based method

The use of human characteristics is the best solution compared to the user that personally knows and possesses [14]. In other words, biometric-based method cannot be forgotten or lost in contrast to token, smart card, and password [24]. A biometric system is basically a pattern recognition system that recognizes a person based on a feature vector derived from a specific physiological or behavioral characteristic that the person possesses or exhibits [25]. These authentication methods identify the user as themselves based on measurable physiological or behavioral characteristics. Physical biometrics

Various technologies of physiological biometrics including finger scan, iris scan, retina scan, hand scan, and facial scan have been proposed and developed using measurements from the human body. There was evidence that the best accuracy can be obtained by using the physical biometric-based method. Table 3 shows a summary of biometric-based method (physical biometric) for user authentication.

[26]Finger scanEdge-based approachEase of useSensitive to camera limitations
[27]Finger scanRule miningGood in case of phone lossBad performance
[28]Finger scanImage preprocessing region segmentationEase of useThe higher templates that save in enrollment database, the execution time for the verification increases
[29]Facial scanVertical pose recovery Fast semi-3D faceExtra security caused by combining with PINHigh energy consumption
[30]Facial scanFragile watermarking based on chaos theoryFast speed of authentication processNot completely secured compared to other techniques
[31]Iris scanDaubechies wavelet transformIncrease the recognition of performance rateTime- and energy-consuming

Table 3.

Summary of biometric-based method (physical biometric).

Fingerprint is the most famous features in biometric-based method and has shown to exhibit the best performance among others. Some of the approaches under fingerprint are of edge-based approach [26] and the rule mining approach [27], as well as the technique of image preprocessing region segmentation [28]. The advantages of using fingerprint are the ease of use and high in authentication accuracy. Nowadays, the fingerprint scanner is used widely among the user.

The concept of facial recognition technique through a vertical pose recovery fast semi-3D face [29] and fragile watermarking based on chaos theory [30] provided an impressive accuracy rate. Moreover, an extra security measure is achievable with the combination of this technique and other user authentication methods such as PIN.

Ref. [31] introduced a Daubechies wavelet transform approach to increase the performance rate for iris recognition. The iris is found to be the most accurate feature and being neither duplicable. Even so, when there are obstacles during the scanning process, the decision on recognition may be disrupted. Behavioral biometric

The other group of biometric-based method is the behavioral biometrics, where users are identified based on their human actions such as signature, gaits (the way humans walk), keystroke dynamics (typing styles), and mouse dynamics [32].

In general, the direction of movement is detected by the magnetometer, while the gait recognition is detected by the gyroscopic sensor and accelerometer [33, 34]. For verification purposes, these authors used the same classifier, which is the K-nearest neighbor (K-NN). The gait recognition has a similar efficiency to the other biometric-based authentication. Nevertheless, the user is required to walk for a certain distance before the process of verification can occur.

Keystroke dynamics is one of the automated methods for verifying the identity of the user based on the manner and rhythm of typing on the keyboard [35]. In paper [36, 37], the authors used the support vector machine (SVM) as a classifier for the development of the system. Another approach that is usually used for the implementation of keystroke dynamics is random forest which can be found in [38].

Signature recognition is another user authentication scheme that works by analyzing handwriting style, in particular the signature. In the offline signature verification, [39] introduced the support vector machine (SVM) classifier, while [40] proposed fuzzy modeling based on the Takagi-Sugeno (TS) model. Table 4 shows a summary of biometric-based method (behavioral biometric) for user authentication.

[33]GaitsLinear regression classifier (KNN)Biometric-based authentication with the same efficiencyDepends on the ideal conditions that the owner holds and operates the device in the same style all the time
[34]GaitsClassifier (KNN)Do not involve explicit user interaction during verification processRequires the punctual calibration of accelerator
[36]Keystroke dynamicsSVMQuick and easy configuration of individual thresholds without impostors’ dataLarge number of data required
[37]Keystroke dynamicsSVMThe cheapest and easiest for the implementation processWasting of time for the user during enrollment process
[38]Keystroke dynamicsRandom forest
  • Low cost

  • Replaceable in the event of compromise

Not sufficient for a high-security environment
[40]SignatureFuzzyWell established for automatic signature verification
  • Limited number of samples to be used for learning

  • The ability of the system to discriminate the forgeries

Table 4.

Summary of biometric-based method (behavioral biometric).

3. Behavioral-based biometric authentication

This section aims to find the good techniques for behavioral-based biometric authentication. Figure 2 shows the various machine learning techniques that can generally be categorized into supervised (classification) and unsupervised (clustering).

Figure 2.

Classification of machine learning techniques.

Supervised machine learning can be used to classify the data much more accurately. In literatures, researchers have used classification techniques such as K-nearest neighbor (K-NN) [41], multilayer perceptron (MLP) [42], dynamic time warping (DTW) [43], neural network [7, 5, 44], decision tree algorithm [45], normalization and leave-one-out method [46], and support vector machine (SVM) [9, 47, 48]. These techniques have improved the performance of the system, and the results have shown some significant achievements in their respective domains. Meanwhile, unsupervised machine learning can be used to perform data reduction task by filtering out unrepresentative data. The data which will not be able to cluster correctly can be considered as outlier’s data. After the reduction task, the classification result is expected to achieve optimal solution. The clustering algorithm can be further subcategorized into flat/partitioning-based and hierarchical-based clustering algorithm [49, 50].

The essential objective for the implementation of the behavioral-based biometric authentication is to acquire the accuracy and also to improve the performance of the system. This goal leads to the creation of a great classifier technique to solve the accuracy problems related to biometric authentication.

Ref. [47] developed an android application using touch-swipe biometric approach. In this work, touchscreen and motion data were collected through a physiological questionnaire. Parameters that are measured were duration, average velocity, mean X, mean Y, mean Z, length of trajectory, acceleration at start, midstrok pressure, midstrok finger area, mean pressure, and mean finger area. The author used support vector machine (SVM) as a classifier, and data analysis was done using WEKA software tool. The result in authentication of equal error rate (EER) was improved from single swipe (4%) to five swipes (0.2%).

Ref. [48] implemented a simple and efficient dynamic user authentication method. Authors also developed the data collection software that runs as the background job and without affecting other applications. This software has extracted the features such as click elapsed time, movement speed, movement acceleration, and relative position of extreme speed and used support vector machine (SVM) technique for classification of the data. This approach achieved the acceptable level of performance with false acceptance rate (FAR) of 0.37% and false rejection rate (FRR) of 1.12%.

Ref. [9] introduced a verification system based on mouse movements using logging tool recording user input (RUI). This system is able to verify a user accurately using newly defined angle-based metrics such as direction, angle of curvature, and curvature distance. This paper used support vector machine (SVM) on the design of the classifier user verification process. Around 30 users participated in this experiment. During their routine computing activities, the mouse movement data were recorded continuously. The result in an EER was recorded at 1.3%.

Ref. [51] used a mouse dynamic dataset from ISOT research lab (University of Victoria). This paper has applied Learning Algorithm for Multivariate Data Analysis (LAMDA) for data analysis. The evaluation of accuracy using 48 users achieved a FAR of 0% and a FRR of 0.36%.

Ref. [6] presented a static approach in which the user needs to perform a task called “follow the maze.” Then, mouse movements are recorded to compute the velocity for X and Y directions. In the verification phase, edit distance (also called Levenshtein distance or dynamic time warping) is used for the purpose of comparison between training and testing dataset. Experiment was conducted involving 28 participants including people highly skilled in computer and people not so skillful in using a mouse device. Nevertheless, they are set to use the same mouse device during the experiment. The result for EER was measured at around 27%.

Ref. [5] presented a continuous user authentication approach with higher-level actions, and the characteristics recorded are distance, action type, direction, and duration. The parameters that are involved in this research were movement speed, direction of movement, type of action, traveled distance, and movement elapsed time. The main experiment involved 22 participants, and 284 hours of raw data are collected over 998 sessions. This paper has applied artificial neural network for the classification of data. The result was presented using receiver operating characteristic (ROC) curves and a confusion matrix yield at the crossover point. This approach achieved the accuracy with FAR of 2.4649% and FRR of 2.4614%.

Ref. [52] proposed a static authentication which presented an enrollment by moving the mouse toward the dots drawn sequentially on the screen. Besides, the user’s mouse movements were computed to generate features for enrolment signature. During verification process, the user follows the dots pattern identical to that of an enrolment phase. Then, this value was compared with the enrollment signature. This experiment involved 15 users, and they must use the same computer and mouse. The equal error rate (EER) for this system was recorded at 15%.

Ref. [45] presented a system that is related to the continuous approach in which raw mouse data was preprocessed to build a model of a user’s behavior. The raw features such as speed, distance, frequency, and angle were extracted to compute the mean, standard deviation, and third-moment values for N data points. This paper has applied a supervised learning method, a decision tree algorithm for classification. This algorithm provides an intelligible representation to discriminate among K users for decision-making process. An authentication experiment was participated by 11 users. They were instructed to run Internet Explorer using their own personal computer. The result achieved for an average false acceptance rate (FAR) was 0.43%, and an average false rejection rate (FRR) was 1.75%.

Ref. [44] introduced an approach for providing secure access over the Internet using biometric authentication. The system used a hybrid approach, which was the combination of keyboard and signature to ensure that the set of credentials supplied to the system at the login stage is genuine. In this experiment, the author developed a web-based applet for the collection of data. For keyboard, the parameters that involved were latency times and hold times, while for signature, the parameters used were angle and distance. This paper was applied in neural network for data analysis. The evaluation of accuracy achieved a FAR of 4.4% and a FRR of 0.2%.

Table 5 shows a list of recent works on different behavioral-based biometric authentication approach that includes the collection of data, the parameter measured, the data analysis, the software used, and the measurement of accuracy. The false rejection rate (FRR), false acceptance rate (FAR), and equal error rate (EER) for every approach are also investigated. Briefly, many classifier techniques have been developed in biometric authentication fields such as neural network, decision tree algorithm, Learning Algorithm for Multivariate Data Analysis (LAMDA), and SVM. However, there is still room to enhance the accuracy of FAR and FRR in this field.

AuthorBiometric approachData collectionParameter (feature extraction)Data analysis (classifier)Software usedMeasurement of accuracy
[47]Touch swipesAndroid (psychological questionnaire)Raw data: touch action; X and Y coordinate; X, Y, and Z gravity; pressure exerted; and finger area Feature vector: duration, length of trajectory, average velocity, acceleration at start, midstrok pressure, midstrok finger area, mean pressure, mean finger area, mean X, mean Y, and mean ZSVMWEKAEER
[48]Mouse dynamicsData collection softwareFeature vector: click elapsed time, movement speed, movement acceleration, and relative position of extreme speedSVMPattern-growth-based mining methodFAR, FRR
[9]Mouse movementRecording user input (RUI)Raw data: action type, time stamp, coordinate X, and coordinate Y Feature vector: three fine-grained angle-based metrics (direction, angle of curvature, and curvature distance)SVMEER
[51]Mouse dynamicsISOT mouse datasetMovement speed, direction of movement, type of action, traveled distance, and movement elapsed timeLearning Algorithm for Multivariate Data Analysis (LAMDA)MATLABFAR, FRR
[6]Mouse dynamicsGUIFeature vector: horizontal and vertical track velocityEdit distance metricsEER
[5]Mouse dynamicsThe client softwareFeature vector: movement speed, direction of movement, type of action, traveled distance, and movement elapsed timeNeural networkMATLABFAR, FRR
[52]Mouse movementGUIFeature vector: speed, deviation, positive angle, and negative angle (average, SD, minimum, maximum)Comparing value with the range of the user’s counter value (exact value)EER
[45]Mouse dynamicsMouse dynamic applicationRaw data: speed, distance, frequency, and angle Feature vector: mean, standard deviation, and third-moment values for N data pointsDecision tree algorithmFAR, FRR
[44]Hybrid approach (keyboard + signature)Web-based appletKeyboard: latency times and hold times Signature: angle and distance (two approaches used to extract—ranking approach and genetic approach)Neural networkFAR, FRR

Table 5.

Comparison of related works for behavioral-based biometric authentication.

4. Discussion

Nowadays, the knowledge-based methods are commonly used because they are simple, economic, and convenient mechanisms to be used and implemented. However, these methods are also known as being an extremely poor form of protection. There are several ways in which an impostor can attack password-protected systems. The most common form of attack is password guessing. Authentication can also use something that user has as alternatives such as tokens, smart card, and QR code. However, these approaches does not lend itself particularly well in the above situation either. These kinds of approaches are more secure to use than a user’s PIN or password. Thus, this possession-based method for user authentication can be considered weaker still. To overcome the drawbacks of those authentication methods, research has been shifted into biometric-based methods for the purposes of authentication, as biometric characteristics are not possible for sharing and repudiating due to uniqueness. Behavioral biometrics is the field of study related to the measure of uniquely identifying measurable patterns in human activities. The term contrasts with physical biometrics, which involves innate human characteristics such as fingerprints or iris patterns. Table 6 shows the user authentication method that can be generally categorized into four categories.

Something the user knowsPIN, password, etc.Can be shared and forgotten
Something the user hasToken, smart card, QR code, etc.Can be lost and duplicated
Something the user isFinger scan, iris scan, retina scan, hand scan, facial scan, etc.Not possible to share and repudiate
Something the user exhibitsSignature, gaits (the way humans walk), keystroke dynamics (typing styles), mouse dynamics, etc.Not possible to share and repudiate

Table 6.

Methodologies of user authentication.

In reality, many behavioral-based biometric methods have been proposed. However, the implementation and deployment are still lacking due to a few reasons such as costly devices, difficult to implement, and sometimes lack of accuracy.

5. Conclusion

This survey provides a comprehensive study on machine learning techniques in the domain of behavioral-based biometric authentication. Particularly, we reassess papers published between the years 2003 and 2016. First, we introduce the concept of biometric authentication and its application. Second, we present the taxonomy of authentication methods with detailed discussion on knowledge-based, possession-based, and biometrics-based methods. In the section of behavioral-based biometric authentication, we discuss the two subcategories of machine learning techniques which are supervised (classification) and unsupervised (clustering) techniques. We investigate each subcategory that has been implemented in the previous behavioral-based biometric authentication. In the end of this paper, we should be able to acquire relevant knowledge required for enhancing the performance of the behavioral-based biometric authentication.


This research was supported by the Malaysian Ministry of Higher Education [Grant No. FRGS/1/2017/ICT03/UNISZA/02/1 (RR228)].

Conflict of interest

All authors agreed that there is no conflict of interests.

© 2018 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite and reference

Link to this chapter Copy to clipboard

Cite this chapter Copy to clipboard

Nurul Afnan Mahadi, Mohamad Afendee Mohamed, Amirul Ihsan Mohamad, Mokhairi Makhtar, Mohd Fadzil Abdul Kadir and Mustafa Mamat (October 31st 2018). A Survey of Machine Learning Techniques for Behavioral-Based Biometric User Authentication, Recent Advances in Cryptography and Network Security, Pinaki Mitra, IntechOpen, DOI: 10.5772/intechopen.76685. Available from:

chapter statistics

408total chapter downloads

More statistics for editors and authors

Login to your personal dashboard for more detailed statistics on your publications.

Access personal reporting

Related Content

This Book

Next chapter

Introductory Chapter: Recent Advances in Cryptography and Network Security

By Pinaki Mitra

Related Book

Frontiers in Guided Wave Optics and Optoelectronics

Edited by Bishnu Pal

First chapter

Frontiers in Guided Wave Optics and Optoelectronics

By Bishnu Pal

We are IntechOpen, the world's leading publisher of Open Access books. Built by scientists, for scientists. Our readership spans scientists, professors, researchers, librarians, and students, as well as business professionals. We share our knowledge and peer-reveiwed research papers with libraries, scientific and engineering societies, and also work with corporate R&D departments and government entities.

More About Us