Keystroke Sequences Research Articles

Robust remote detection of depressive tendency based on keystroke dynamics and behavioural characteristics

Depressive Disorder (DD) is a leading cause of disability worldwide. Screening tools for detecting DD symptoms are essential for monitoring and efficient managing. Remarkably, individuals’ kinetic activities, including their interaction with touchscreen smartphones, can be a proxy for their mental status. Therefore, studying these typing patterns can assist in developing passive screening tools for detecting even the early stage of DD, i.e., the depressive tendency (DT). Here we extend a previous study by exploring different machine learning models with various feature engineering approaches to detect the subjects’ DT, as indicated by the self-administered Patient Health Questionnaire-9 (PHQ-9) score, via keystroke digital biomarkers. The keystroke timing sequences were unobtrusively collected from 24 subjects during routine interaction with touchscreen smartphones, resulting in 23,264 typing sessions. The proposed framework was investigated under two keystroke feature combinations—hold-time and flight-time variables—and validated using nested cross-validation scheme. Different feature selection (FS) techniques were employed to select informative features from the keystroke sequences. The best-performing gradient boosting classifier with features selected by the mutual information FS method achieved an improved Area Under Curve (AUC) of 0.98 [95% confidence interval: 0.91-1.00]. The proposed DT pipeline, which surpasses the state-of-the-art models, could effectively capture DT, considering users’ behavioural characteristics. This would potentially provide users with information regarding the evolution of their mental health, simultaneously contributing to improving digital tools for objectively screening mental disorders in-the-wild.

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.

Investigating Problem Solving on Calculator Items in a Large-Scale Digitally Based Assessment: A Data Mining Approach

This exploratory study investigated the behaviors and content of onscreen calculator usage by a nationally representative sample of eighth-grade students who responded to items from the 2017 National Assessment of Educational Progress mathematics assessment. Meaningful features were generated from the process data to infer whether students spontaneously used calculators for mathematical problem solving, how frequently and when they used them, and the nature of the operations performed on calculators. Sequential pattern mining was applied on sequences of calculator keystrokes to obtain patterns of operations that were representative of students’ problem-solving strategies or processes. Results indicated that higher scoring students not only were more likely to use calculators, but also used them in a more goal-driven manner than lower scoring students.

Using sequence mining to study students’ calculator use, problem solving, and mathematics achievement in the National Assessment of Educational Progress (NAEP)

Using appropriate tools strategically to aid in problem solving is a crucial skill identified in K-12 mathematics curriculum standards. As more assessments transition from paper-and-pencil to digital formats, a variety of interactive tools have been made available to test takers in digital testing platforms. Using onscreen calculators as an example, this study illustrates how process data obtained from student interactions with a digitally-based large-scale assessment can be leveraged to explore how and how well test takers use interactive tools and unveil their mathematical problem-solving processes and strategies. Specifically, sequence mining techniques using the longest common subsequence were applied on process data collected from a nationally representative sample who took the National Assessment of Educational Progress (NAEP) mathematics assessment to examine patterns of eighth-grade students’ calculator-use behaviors and the content of calculator input across a series of items. Sequences of keystrokes executed on the onscreen calculator by test takers were compared to reference sequences identified by content experts as proficient and efficient use to infer how well and how consistently the calculator was used. Results indicated that calculator-use behaviors and content differed by item characteristics. Students were more likely to use calculators on calculation-demanding items that involve intensive and complex computations than on items that involve simple or no computation. Using the calculator on more calculation-demanding items and using it in a manner that is more efficient and more similar to reference sequences on these items were related to higher mathematical proficiency. Findings have implications for assessment design and can be used in educational practices to provide educators with actionable process-related information on tool use and problem solving.

TypeNet: Deep Learning Keystroke Biometrics

We study the performance of Long Short-Term Memory networks for keystroke biometric authentication at large scale in free-text scenarios. For this we explore the performance of Long Short-Term Memory (LSTMs) networks trained with a moderate number of keystrokes per identity and evaluated under different scenarios including: i) three learning approaches depending on the loss function (softmax, contrastive, and triplet loss); ii) different number of training samples and lengths of keystroke sequences; iii) four databases based on two device types (physical vs touchscreen keyboard); and iv) comparison with existing approaches based on both traditional statistical methods and deep learning architectures. Our approach called TypeNet achieves state-of-the-art keystroke biometric authentication performance with an Equal Error Rate of 2.2% and 9.2% for physical and touchscreen keyboards, respectively, significantly outperforming previous approaches. Our experiments demonstrate a moderate increase in error with up to 100,000 subjects, demonstrating the potential of TypeNet to operate at an Internet scale. To the best of our knowledge, the databases used in this work are the largest existing free-text keystroke databases available for research with more than 136 million keystrokes from 168,000 subjects in physical keyboards, and 60,000 subjects with more than 63 million keystrokes acquired on mobile touchscreens.

“You're wrong, I'll switch, I'm wrong, I'll stay”: How task-switching strategies are modulated by a partner in a multi-task learning protocol

Individuals given control over practice variables make practice decisions based on their current performance. When individuals practice in pairs, the question as to if and how a partner's performance impacts these decisions is of theoretical and practical interest. Here, we evaluated this question in a multi-task learning protocol, where individuals and dyads practiced three, differently timed keystroke sequences. Dyad participants alternated turns with a partner so we could study the immediate consequences of the partner's performance on practice choice. Only one of the partners had choice over the sequence order, the other partner practiced the sequences in either a predetermined blocked or random order. Practice with a partner that had a random-schedule promoted more task-switching in the other partner and had some benefit for retention accuracy. Distinct “own-error” and “partner-error” switching strategies were evidenced, with partners choosing to repeat the same sequence on their next turn when they performed poorly or when their partner performed well. These data show that an individual's practice decisions are influenced by their social context, particularly the practice schedule and patterns of errors in a partner's performance.

TKCA: a timely keystroke-based continuous user authentication with short keystroke sequence in uncontrolled settings

Keystroke-based behavioral biometrics have been proven effective for continuous user authentication. Current state-of-the-art algorithms have achieved outstanding results in long text or short text collected by doing some tasks. It remains a considerable challenge to authenticate users continuously and accurately with short keystroke inputs collected in uncontrolled settings. In this work, we propose a Timely Keystroke-based method for Continuous user Authentication, named TKCA. It integrates the key name and two kinds of timing features through an embedding mechanism. And it captures the relationship between context keystrokes by the Bidirectional Long Short-Term Memory (Bi-LSTM) network. We conduct a series of experiments to validate it on a public dataset - the Clarkson II dataset collected in a completely uncontrolled and natural setting. Experiment results show that the proposed TKCA achieves state-of-the-art performance with 8.28% of EER when using only 30 keystrokes and 2.78% of EER when using 190 keystrokes.

WiPass: 1D-CNN-based smartphone keystroke recognition Using WiFi signals

Keystroke privacy is crucial for smartphone system security and user privacy. User’s private information may be leaked if the sequences of keystrokes on the numeric soft keyboard are obtained by an adversary in a certain way. Different keystrokes may lead to different finger movements, causing diverse interference to WiFi signals, which can be indicated by the fluctuation of Channel State Information (CSI) waveforms. In this paper, we propose WiPass, a keystroke recognition system for classifying numeric keyboard inputs on smartphones, which consists of a transmitter (e.g. a WiFi router) and a receiver (e.g. a desktop computer with a Commercial Off-The-Shelf (COTS) WiFi NIC). The key inspiration comes from the fact that while performing a certain keystroke near a receiver, the CSI values received by the receiver vary in a unique pattern. WiPass can extract and analyze the CSI Data generated by user’s keystroke operations on the smartphone, thus inferring the users’ numeric keystrokes by comparing and classifying the CSI waveforms of the different keystrokes. Distinct from the previous keystroke inference approaches, WiPass employs 1D Convolutional Neural Network (1D-CNN) model as the classification model instead of other machine learning models. The experimental results show that the accuracy rate of the WiPass in detecting a keystroke and in classifying keystrokes reaches over 95% and 85% respectively.

Continuous authentication by free-text keystroke based on CNN and RNN

Personal keystroke modes are difficult to imitate and can therefore be used for identity authentication. The keystroke habits of a person can be learned according to the keystroke data generated when the person inputs free text. Detecting a user's keystroke habits as the user enters text can continuously verify the user's identity without affecting user input. The method proposed in this paper authenticates users via their keystrokes when they type free text. The user keystroke data is divided into a fixed-length keystroke sequence, which is then converted into a keystroke vector sequence according to the time feature of the keystroke. A model that combines a convolutional neural network and a recursive neural network is used to learn a sequence of individual keystroke vectors to obtain individual keystroke features for identity authentication. The model is tested using two open datasets, and the best false rejection rate (FRR) is found to be (2.07%,6.61%), the best false acceptance rate (FAR) is found to be (3.26%, 5.31%), and the best equal error rate (EER) is found to be (2.67%, 5.97%).

Lexical and Frequency Effects on Keystroke Timing: Challenges to a Lexical Search Account From a Type-To-Copy Task

We explore how timing in identical keystroke sequences of a stem morpheme is influenced by linguistic knowledge by manipulating lexical status and morphological complexity of words in a type-to-copy task. Starting from the second keystroke, we find that average keystroke latency within a stem morpheme varies according to whole-word frequency (Experiment1) and lexicality defined by compatibility of the upcoming suffix (e.g., IZE vs IST) with the stem (e.g., NORMAL) to form a string (e.g., RENORMALIZE vs. RENORMAL; RENORMALIZE vs. RENORMALIST in Experiments 2 & 3 respectively). Further, although lexical and frequency effects persist over the string as a whole, nonlinear mixed-effects regressions reveal position varying lexical effects on keystroke latencies within the stem morpheme. In addition, whole word frequency effects on the first keystroke were present. These results challenge hierarchical accounts of production with modular motor programs where the same letter sequence (for a morpheme) is realized independently of and only after lexical access to the word in which the letters occur (cf. Crump & Logan, 2010; Logan & Crump 2011).

Tracking Keystroke Sequences at the Cortical Level Reveals the Dynamics of Serial Order Production.

Response selection is often studied by examining single responses, although most actions are performed within an overarching sequence. Understanding processes that order and execute items in a sequence is thus essential to give a complete picture of response selection. In this study, we investigate response selection by comparing single responses and response sequences as well as unimanual and bimanual sequences. We recorded EEG while participants were typing one- or two-keystroke sequences. Irrespective of stimulus modality (visual or auditory), response-locked analysis revealed distinct contralateral and ipsilateral components previously associated with activation and inhibition of alternative responses. Unimanual sequences exhibited a similar activation/inhibition pattern as single responses, but with the activation component of the pattern expressed more strongly, reflecting the fact that the hand will be used for two strokes. In contrast, bimanual sequences were associated with successive activation of each of the corresponding motor cortices controlling each keystroke and no traceable inhibitory component. In short, the activation component of the two-keystroke sequence EEG pattern can be understood from the addition of activation components of single-stroke sequences; the inhibition of the hand not being used is only evidenced when that hand is not planned for the next stroke.

Analysis of Keystroke Sequences in Writing Logs

AbstractThe research on using event logs and item response time to study test‐taking processes is rapidly growing in the field of educational measurement. In this study, we analyzed the keystroke logs collected from 761 middle school students in the United States as they completed a persuasive writing task. Seven variables were extracted from the keystroke logs and compared with different score and gender groups. Group comparisons were also made using methodologies borrowed from sequence mining. Students' composition strategies over the course of the writing process were also investigated. The findings of this study have implications for gaining deeper understanding of observed group differences and for designing interventions to close the achievement gaps among population groups.

Continuous authentication by free-text keystroke based on CNN plus RNN

Personal keystroke mode is difficult to imitate and can therefore be used for identity authentication. According to the keystroke data when a person inputs free text, the keystroke habit of the person can be learned. Detecting a user’s keystroke habits as the user enters text can continuously verify the user’s identity without affecting user input. This paper proposes to divide the user keystroke data into a fixed-length keystroke sequence, and convert the keystroke sequence into a keystroke vector sequence according to the time feature of the keystroke. A model of a recursive neural network plus a convolutional neural network is used to learn a sequence of individual keystroke vectors to obtain individual keystroke features for identity authentication. The model was tested using an open data set and the best False Rejection Rate(FRR) was 1.95%, False Acceptance Rate (FAR) was 4.12% and Equal Error Rate(EER) was 3.04%.

Erratum to: Measuring sequences of keystrokes with jsPsych: Reliability of response times and interkeystroke intervals.

Erratum to: Measuring sequences of keystrokes with jsPsych: Reliability of response times and interkeystroke intervals.

Measuring sequences of keystrokes with jsPsych: Reliability of response times and interkeystroke intervals.

Although the precision and reliability of response time (RT) measurements performed via Web-based interfaces have been evaluated, sequences of keystrokes have never been investigated in detail. Keystrokes often occur in much more rapid succession than RTs, and operating systems may interpret successive or concomitant keyboard events according to both automatic and user-based settings. Sequence keystroke timing could thus be more sensitive than single RTs to noise in online measurements. Here, we quantified the precision and reliability of timing measures performed during sequences of keystrokes. We used the JavaScript jsPsych library to create an experiment involving finger-movement sequences, and ran it online with 633 participants. We manipulated the structure of three keystroke motor sequences, targeting a replication of previous findings regarding both RTs and interkeystroke intervals (IKIs). Our online data accurately reproduced the original results and allowed for a novel assessment of demographic variables such as age and gender. In parallel, we also measured the objective timing accuracy of the jsPsych interface by using specialized hardware and software, showing a constant 60-ms delay for RTs and a 0-ms delay for IKIs across the sequences. The distribution of IKIs revealed quantizing for a majority of participants, most likely due to the sampling frequency of their USB keyboards. Overall, these findings indicate that JsPsych provides good reliability and accuracy in sequence keystroke timings for mental chronometry purposes, through online recordings.

Keystroke Biometrics Ongoing Competition

This paper presents the first Keystroke Biometrics Ongoing Competition (KBOC) organized to establish a reproducible baseline in person authentication using keystroke biometrics. The competition has been developed using the BEAT platform and includes one of the largest keystroke databases publicly available based on a fixed text scenario. The database includes genuine and attacker keystroke sequences from 300 users acquired in four different sessions distributed in a four month time span. The sequences correspond to the user’s name and surname, and therefore, each user comprises an individual and personal sequence. As baseline for KBOC, we report the results of 31 different algorithms evaluated according to accuracy and robustness. The systems have achieved EERs as low as 5.32% and high robustness to multisession variability with accuracy degradation lower than 1% for probes separated by months. The entire database is publicly available at the competition website.

EFFECTS OF REPETITION FREQUENCY ON OPERANT STRENGTH AND RESURGENCE OF NON-CRITERIAL FEATURES OF OPERANTS

The present research is based on the conceptualization of resurgence as reappearance of behavior that occurred earlier in the individual’s history but not recently, without restoration of the conditions under which the earlier behavior occurred. In a series of five experiments, human participants typed nonword sequences of letters on a computer keyboard. Each sequence was initiated with the spacebar and ended with the enter key, and was treated as a “revealed operant.” Each operant was composed of criterial (mandated) and noncriterial (discretionary) keystrokes. Participants learned several unique operants, each defined by a different set of criterial keystrokes. The objective was to study the effect of varying the number of repetitions required for each operant during the learning sessions on the relative frequency with which those operants were performed during a test session. The operants that had previously been performed most frequently were chosen for performance most often. Noncriterial resurgence was measured by “antiquity”—how far back one has to go in the participant’s prior history to find previous instances of that same noncriterial keystroke sequence. Criterial and noncriterial components of the operants were affected by the independent variable in different ways.

English

Biometrics is an automated method of recognizing a person based on a physiological or behavioral appearance. Biometric detection technology relies upon the physical appearance of an individual, such as fingerprints, voiceprint, prototype of the iris of the eye and facial prototype, in identifying an individual, offering positive identification that is difficult to counterfeit. Instances of physiological biometric traits include height, weight, body odour, the aspect of the hand, the prototype of veins, retina or iris, the face and the prototypes on the skin of thumbs or fingers (fingerprints). Instances of behavioral biometrics are voice prototypes, signature and keystroke sequences and gait (the body movement while walking). Biometric identifiers are the distinctive, measurable appearances used to label and describe individuals [1]. Currently, passwords, Personal Identification cards are used for personal detection. However, cards can be stolen, and passwords and numbers can be guessed or forgotten. To solve these problems, biometric verification technology, which identifies people by their unique biological information, is attracting attention [2]. Biometrics is mostly used for authentication purposes to ensure proper safety.

Evaluating the Effectiveness of Peer-Scheduled Practice on Motor Learning

Giving learners a choice over how to schedule practice benefits motor learning. Here we studied peer scheduling to determine whether this benefit is related to the adaptive nature of practice or decisions about how to switch between skills. Forty-eight participants were paired and assigned to self- or peer-scheduled groups. Within each pair, one person (Actor) physically practiced 3 keystroke sequences, each with different timing goals. Self-scheduled Actors chose the sequence before each practice trial while their Partner watched. Peer-scheduled Actors had their practice directed by their Partner. Both peer schedulers and self-schedulers showed performance-dependent practice, making decisions to switch based on timing error. However, peer schedulers generally chose to switch more than self-schedulers although this was not related to retention for either group. Importantly, self-scheduled Actors did not differ in retention from peer-scheduled Actors, but the Actors generally performed with lower error in retention than that of their partners. Peer-scheduled practice was rated as more motivating and enjoyable than self-scheduled practice. In view of the lack of difference in retention and the positive ratings of peer-scheduled practice, we conclude that it is the adaptive nature of practice that is important for learning and that peer-directed practice is an effective alternative practice method to self-directed practice.

The problem of serial order in skilled typing.

We address the problem of serial order in skilled typing, asking whether typists represent the identity and order of the keystrokes they type jointly by linking successive keystrokes into a chained sequence, or separately by associating keystrokes with position codes. In 4 experiments, typists prepared to type a prime word and were probed to type a target word. We varied the overlap between the identity and order of keystrokes in the prime and the target. Experiment 1 tested whether the identity of keystrokes can be primed separately from their order. Experiments 2 and 3 tested whether keystroke positions can be primed out of sequence. Experiment 4 tested whether keystrokes are primed equally across serial positions. The results were consistent with chaining theories: Keystroke identities were not primed separately from their order, keystroke positions were not primed out of sequence, and priming was graded across the keystroke sequence and depended on the number of keystrokes that were primed in sequence. We conclude by discussing the possibility that the problem of serial order may be solved differently for different sequential tasks.