Applied Soft Computing 34 (2015) 178–193 Contents lists available at ScienceDirect Applied Soft Computing j ourna l ho me page: www.elsevier.com/locate /asoc Emphasizing typing signature in keystroke dynamics using immune algorithms Paulo Henrique Pisani a,∗ , Ana Carolina Lorena b a Universidade Federal do ABC (UFABC), Centro de Matemática, Computac ¸ ão e Cognic ¸ ão (CMCC), Av. dos Estados, 5001, Santo André, Brazil b Universidade Federal de São Paulo (UNIFESP), Instituto de Ciência e Tecnologia (ICT), Rua Talim, 330, São José dos Campos, Brazil a r t i c l e i n f o Article history: Received 24 January 2014 Received in revised form 11 January 2015 Accepted 10 May 2015 Available online 16 May 2015 Keywords: One-class classification Data pre-processing Immune algorithms Keystroke dynamics a b s t r a c t Improved authentication mechanisms are needed to cope with the increased data exposure we face nowadays. Keystroke dynamics is a cost-effective alternative, which usually only requires a standard keyboard to acquire authentication data. Here, we focus on recognizing users by keystroke dynamics using immune algorithms, considering a one-class classification approach. In such a scenario, only samples from the legitimate user are available to generate the model of the user. Throughout the paper, we emphasize the importance of proper data understanding and pre-processing. We show that keystroke samples from the same user present similarities in what we call typing signature. A proposal to take advantage of this finding is discussed: the use of rank transformation. This transformation improved performance of classification algorithms tested here and it was decisive for some immune algorithms studied in our setting. © 2015 Elsevier B.V. All rights reserved. 1. Introduction The current technological scenario has brought a number of improved services to society, particularly owing to Internet-based applications. However, at the same time, this scenario has con- tributed to increase data exposure, giving a new momentum to concerns regarding identity theft. Thereby, there is a need to enhanced authentication mechanisms. A possible alternative is by the use of biometrics. In security area, biometrics tries to recognize users by physiological or behavioral features of the person. There are several biometrics technologies currently available. This work focuses on keystroke dynamics, which studies ways to recognize users by their typing rhythm. This technology shows as being a promising alternative due to several reasons [1,2]. Firstly, it usually does not need any additional cost with hardware, as a common keyboard is enough to acquire keystroke data. Other bio- metric technologies, such as fingerprint or iris recognition, require a specific device to acquire biometric data. Secondly, keystroke dynamics recognition may be performed in background, while the user is typing an e-mail or entering a password. Consequently, day- ∗ Corresponding author. Present address: Universidade de São Paulo (USP), Instituto de Ciências Matemáticas e de Computac ¸ ão (ICMC), Av. Trabalhador São- carlense, 400, São Carlos, Brazil. Tel.: +55 163373 9700. E-mail addresses: phpisani@icmc.usp.br (P.H. Pisani), aclorena@unifesp.br (A.C. Lorena). to-day tasks are not disturbed, what may contribute to a better acceptability of the technology by the user. A keystroke dynamics system should be able to distinguish a legitimate user from potential intruders, a classic binary clas- sification setting in pattern recognition and machine learning. Nonetheless, collecting intruders data can be impractical in day- to-day use of computational systems. This makes a one-class classification setting more appropriate, where the user model is built using data from the legitimate user only. Several algorithms have been applied for classifying users by keystroke dynamics, in both one-class and conventional two-class settings [3–5]. This paper focuses on immune algorithms, which attained good performance in some of our previous works [6–8]. In this paper we perform a deeper analysis of immune systems in the context of keystroke dynamics, analysing its performance under various aspects. The main goals are: • Show that proper data understanding and preprocessing can be crucial in keystroke dynamics. • Apply rank transformation in keystroke dynamics in order to improve recognition performance. This transformation can emphasize what is called here as typing signature. Dealing with keystroke dynamics requires proper data under- standing and preprocessing, as in the case of other areas [9,10]. Without it, classification algorithms may fail to reach optimal http://dx.doi.org/10.1016/j.asoc.2015.05.008 1568-4946/© 2015 Elsevier B.V. All rights reserved.