Free-text Keystroke Dynamics Research Articles

Cross-device free-text keystroke dynamics authentication using federated learning

Cross-device free-text keystroke dynamics authentication using federated learning

TypeFormer: transformers for mobile keystroke biometrics

The broad usage of mobile devices nowadays, the sensitiveness of the information contained in them, and the shortcomings of current mobile user authentication methods are calling for novel, secure, and unobtrusive solutions to verify the users’ identity. In this article, we propose TypeFormer, a novel transformer architecture to model free-text keystroke dynamics performed on mobile devices for the purpose of user authentication. The proposed model consists in temporal and channel modules enclosing two long short-term memory recurrent layers, Gaussian range encoding, a multi-head self-attention mechanism, and a block-recurrent transformer layer. Experimenting on one of the largest public databases to date, the Aalto mobile keystroke database, TypeFormer outperforms current state-of-the-art systems achieving equal error rate values of 3.25% using only five enrolment sessions of 50 keystrokes each. In such way, we contribute to reducing the traditional performance gap of the challenging mobile free-text scenario with respect to its desktop and fixed-text counterparts. To highlight the design rationale, an analysis of the experimental results of the different modules implemented in the development of TypeFormer is carried out. Additionally, we analyse the behaviour of the model with different experimental configurations such as the length of the keystroke sequences and the amount of enrolment sessions, showing margin for improvement.

Impact of using different-sized touch keyboards on free-text keystroke dynamics authentication in the Arabic language

Authentication using keystroke dynamics (KD) has become an active research area due to its usability and security aspects. Nowadays, the scale of touch keyboard use has expanded to include most modern devices. Although KD typically focuses on a single device at a particular time, authentication systems are adjustable within their environments, as these systems’ users frequently switch between multiple devices. Thus, this paper assesses users’ typing behavior on different tablet devices with varying touch keyboard sizes. In particular, we empirically assess whether the validation results from free-text KD authentication vary depending on the touch keyboard size. The results reveal interesting research directions for future feasibility studies on changing the dynamic keystroke typing-pad effect on user-security and trust-authentication analysis.

On the shape of timings distributions in free-text keystroke dynamics profiles

Keystroke dynamics is a soft biometric trait. Although the shape of the timing distributions in keystroke dynamics profiles is a central element for the accurate modeling of the behavioral patterns of the user, a simplified approach has been to presuppose normality. Careful consideration of the individual shapes for the timing models could lead to improvements in the error rates of current methods or possibly inspire new ones. The main objective of this study is to compare several heavy-tailed and positively skewed candidate distributions in order to rank them according to their merit for fitting timing histograms in keystroke dynamics profiles. Results are summarized in three ways: counting how many times each candidate distribution provides the best fit and ranking them in order of success, measuring average information content, and ranking candidate distributions according to the frequency of hypothesis rejection with an Anderson-Darling goodness of fit test. Seven distributions with two parameters and seven with three were evaluated against three publicly available free-text keystroke dynamics datasets. The results confirm the established use in the research community of the log-normal distribution, in its two- and three-parameter variations, as excellent choices for modeling the shape of timings histograms in keystroke dynamics profiles. However, the log-logistic distribution emerges as a clear winner among all two- and three-parameter candidates, consistently surpassing the log-normal and all the rest under the three evaluation criteria for both hold and flight times.

On the Shape of Timing Distributions in Free-Text Keystroke Dynamics Profiles

Keystroke dynamics is a soft biometric trait. Although the shape of the timing distributions in keystroke dynamics profiles is a central element for the accurate modeling of the behavioral patterns of the user, a simplified approach has been to presuppose normality. Careful consideration of the individual shapes for the timing models could lead to improvements in the error rates of current methods or possibly inspire new ones. The main objective of this study is to compare several heavy-tailed and positively skewed candidate distributions in order to rank them according to their merit for fitting timing histograms in keystroke dynamics profiles. Results are summarized in three ways: counting how many times each candidate distribution provides the best fit and ranking them in order of success, measuring average information content, and ranking candidate distributions according to the frequency of hypothesis rejection with an Anderson-Darling goodness of fit test. Seven distributions with two parameters and seven with three were evaluated against three publicly available free-text keystroke dynamics datasets. The results confirm the established use in the research community of the log-normal distribution, in its two- and three-parameter variations, as excellent choices for modeling the shape of timings histograms in keystroke dynamics profiles. However, the log-logistic distribution emerges as a clear winner among all two- and three-parameter candidates, consistently surpassing the log-normal and all the rest under the three evaluation criteria for both hold and flight times.

Investigating the accuracy of free-text keystroke dynamics authentication in touchscreen devices

Security on smartphones has become a substantial topic since the amount of security relevant data and information stored on them has been increasing. While the password is commonly used as a traditional approach for user authentication, it suffers from a security-usability trade-off dilemma. Accordingly, the keystroke dynamics authentication approach can provide ease of use to the user as well as robust security. This paper investigates the feasibility of utilising user-typing behaviours on touchscreen keypads for the authentication process via free-text keystroke dynamics. In constructing the timing vectors, three timing features are used: hold time, flight time and digraph duration. Additionally, Euclidian and Manhattan distances were utilised to find the degree of similarity between the user's login data and user's profile. An Android application was developed and evaluated through an experimental study. Thus, we achieved a very encouraging result, as we were able to show that applying the free-text method indeed influences the accuracy of user authentication on a touchscreen device.

Investigating the accuracy of free-text keystroke dynamics authentication in touchscreen devices

Security on smartphones has become a substantial topic since the amount of security relevant data and information stored on them has been increasing. While the password is commonly used as a traditional approach for user authentication, it suffers from a security-usability trade-off dilemma. Accordingly, the keystroke dynamics authentication approach can provide ease of use to the user as well as robust security. This paper investigates the feasibility of utilising user-typing behaviours on touchscreen keypads for the authentication process via free-text keystroke dynamics. In constructing the timing vectors, three timing features are used: hold time, flight time and digraph duration. Additionally, Euclidian and Manhattan distances were utilised to find the degree of similarity between the user's login data and user's profile. An Android application was developed and evaluated through an experimental study. Thus, we achieved a very encouraging result, as we were able to show that applying the free-text method indeed influences the accuracy of user authentication on a touchscreen device.

The impact of using different keyboards on free-text keystroke dynamics authentication for Arabic language

PurposeNowadays, there is a high demand for online services and applications. However, there is a challenge to keep these applications secured by applying different methods rather than using the traditional approaches such as passwords and usernames. Keystroke dynamics is one of the alternative authentication methods that provide high level of security in which the used keyboard plays an important role in the recognition accuracy. To guarantee the robustness of a system in different practical situations, there is a need to examine how much the performance of the system is affected by changing the keyboard layout. This paper aims to investigate the impact of using different keyboards on the recognition accuracy for Arabic free-text typing.Design/methodology/approachTo evaluate how much the performance of the system is affected by changing the keyboard layout, an experimental study is conducted by using two different keyboards which are a Mac’s keyboard and an HP’s keyboard.FindingsBy using the Mac’s keyboard, the results showed that the false rejection rate (FRR) was 0.20, whilst the false acceptance rate (FAR) was 0.44. However, these values have changed when using the HP’s keyboard where the FRR was equal to 0.08 and the FAR was equal to 0.60.Research limitations/implicationsThe number of participants in the experiment, as the authors were targeting much more participants.Originality/valueThese results showed for the first time the impact of the keyboards on the system’s performance regarding the recognition accuracy when using Arabic free-text.

Improving the performance of free-text keystroke dynamics authentication by fusion

Free-text keystroke dynamics is invariably hampered by the huge amount of data needed to train the system. This problem has been addressed in this paper by suggesting a system that combines two methods, both of which provide a reduced training requirement for user authentication using free-text keystrokes. The two methods were fused to achieve error rates lower than those produced by each method separately. Two fusion schemes, namely: decision-level fusion and feature-level fusion, were applied. Feature-level fusion was done by concatenating two sets of features before the learning stage. The two sets of features were: a timing feature set and a non-conventional feature set. Moreover, decision-level fusion was used to merge the output of two methods using majority voting. One is Support Vector Machines (SVMs) together with Ant Colony Optimization (ACO) feature selection and the other is decision trees (DTs). Even though the classifiers using the parameters merged at feature level produced low error rates, its results were outperformed by the results achieved by the decision-level fusion scheme. Decision-level fusion was employed to achieve the best performance of 0.00% False Accept Rate (FAR) and 0.00% False Reject Rate (FRR).

Non-conventional keystroke dynamics for user authentication

This paper introduces an approach for user authentication using free-text keystroke dynamics which incorporates the use of non-conventional keystroke features. Semi-timing features along with editing features are extracted from the user’s typing stream. Decision trees were exploited to classify each of the user’s data. In parallel for comparison, support vector machines (SVMs) were also used for classification in association with an ant colony optimization (ACO) feature selection technique. The results obtained from this study are encouraging as low false accept rates (FAR) and false reject rates (FRR) were achieved in the experimentation phase. This signifies that satisfactory overall system performance was achieved by using the typing attributes in the proposed approach. Thus, the use of non-conventional typing features improves the understanding of human typing behavior and therefore, provides significant contribution to the authentication system.

Free‐text keystroke dynamics authentication for Arabic language

This study introduces an approach for user authentication using free-text keystroke dynamics which incorporates text in Arabic language. The Arabic language has completely different characteristics to those of English. The approach followed in this study involves the use of the keyboard's key-layout. The method extracts timing features from specific key-pairs in the typed text. Decision trees were exploited to classify each of the users' data. In parallel for comparison, support vector machines were also used for classification in association with an ant colony optimisation feature selection technique. The results obtained from this study are encouraging as low false accept rates and false reject rates were achieved in the experimentation phase. This signifies that satisfactory overall system performance was achieved by using the typing attributes in the proposed approach, while typing Arabic text.

Biometric Recognition Based on Free-Text Keystroke Dynamics

Accurate recognition of free text keystroke dynamics is challenging due to the unstructured and sparse nature of the data and its underlying variability. As a result, most of the approaches published in the literature on free text recognition, except for one recent one, have reported extremely high error rates. In this paper, we present a new approach for the free text analysis of keystrokes that combines monograph and digraph analysis, and uses a neural network to predict missing digraphs based on the relation between the monitored keystrokes. Our proposed approach achieves an accuracy level comparable to the best results obtained through related techniques in the literature, while achieving a far lower processing time. Experimental evaluation involving 53 users in a heterogeneous environment yields a false acceptance ratio (FAR) of 0.0152% and a false rejection ratio (FRR) of 4.82%, at an equal error rate (EER) of 2.46%. Our follow-up experiment, in a homogeneous environment with 17 users, yields FAR=0% and FRR=5.01%, at EER=2.13%.