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

Abstract

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.

Leveraging Response Times in Learning Environments: Opportunities and Challenges

“…This type of research has been done particularly in the context of cognitive science, where the fitted distributions are connected with the (hypothesized) cognitive processes that generate response times (De Boeck and Jeon, 2019;Van Zandt, 2000;Ratcliff and Rouder, 1998). The exact distribution of response times was also analyzed in the case of keystroke timings, where the motivation is the use of response times as a biometric trait (González et al, 2021). In psychomet-rics, fitting response time distributions often involves mixture modeling (Lee and Chen, 2011).…”

Generalización del Modelado de Cadencias de Tecleo con Contextos Finitos para su Utilización en Ataques de Presentación y Canal Lateral

“…Por un lado, aunque estas distribuciones tienden a ser sesgadas a la derecha y de cola larga, no siempre lo son y, al mostrar irregularidades y picos inconsistentes, pueden diferir significativamente de sus aproximaciones suaves y que decrecen mon ótonamente hacia la derecha. No solo son las distribuciones empíricas raramente gaussianas, sino que incluso fallan una prueba de hip ótesis para el mejor ajuste contra una log-normal entre el 10 % y el 20 % de las veces, como se mostrará en la secci ón 6.1 [19]. Por otro lado, no esperamos que las muestras individuales bajo evaluaci ón contengan suficientes observaciones de tiempo para cada tecla y contexto como para construir un histograma lo suficientemente detallado que permita hacer un test de hip ótesis.…”

“…En lugar de presuponer una cierta distribuci ón suave para los tiempos de retenci ón y latencia y estimar sus paráme-tros en base al perfil intrausuario, se propuso utilizar la totalidad de las muestras en los histogramas empíricos y calcular las distancias correspondientes en base a sus formas particulares para cada usuario y cada digrama. Las distancias basadas en histogramas empíricos han sido descritas en la secci ón A.6, y han sido utilizadas en los artículos [19], [13], y [14].…”

“…También se simplifica la expresi ón pues E[F S i ] = 0,5 (A. 19) de donde la expresi ón para Z-score basado en CDF resulta…”