Expert Systems With Applications 165 (2021) 113967 Available online 5 September 2020 0957-4174/© 2020 Elsevier Ltd. All rights reserved. Contents lists available at ScienceDirect Expert Systems With Applications journal homepage: www.elsevier.com/locate/eswa Analysis of factors that influence the performance of biometric systems based on EEG signals Dustin Carrión-Ojeda a , Rigoberto Fonseca-Delgado b , Israel Pineda a,∗ a School of Mathematical and Computational Sciences, Yachay Tech University, Urcuquí, Ecuador b Department of Computer Sciences, National Institute of Astrophysics, Optics and Electronics, Sta. Ma. Tonantzintla, Mexico ARTICLE INFO Keywords: Biometrics Electroencephalogram Discrete Wavelet Transform Performance factors ABSTRACT Searching for new biometric traits is currently a necessity because traditional ones such as fingerprint, voice, or face are highly prone to forgery. For this reason, the study of bioelectric signals has the potential to develop new biometric systems. A motivation for using electroencephalogram signals is that they are unique to each person and are much more difficult to replicate than conventional biometrics. The objective of this study is to analyze the factors that influence the performance of a biometric system based on electroencephalogram signals. This work uses six different classifiers to compare several decomposition levels of the discrete wavelet transform as a preprocessing technique and also explores the importance of the recording time. These classifiers are Gaussian Naïve Bayes Classifier, K-Nearest Neighbors, Random Forest, AdaBoost, Support Vector Machine, and Multilayer Perceptron. This work proves that the decomposition level does not have a high impact on the overall result of the system. On the other hand, the recording time of electroencephalograms has a significant impact on the performance of the classifiers. It is worth mentioning that this study used two different datasets to validate the results. Finally, our experiments show that Support Vector Machine and AdaBoost are the best classifiers for this particular problem since they achieved a sensitivity, specificity, and accuracy of 85.94 ± 1.8, 99.55 ± 0.06, 99.12 ± 0.11 and 95.54 ± 0.53, 99.91 ± 0.01, and 99.83 ± 0.02 respectively. 1. Introduction With the rapid development of technologies, biometric systems are present in many daily scenarios to provide security to data (Man- zoor & Selwal, 2018). Nowadays, most smartphones have at least one embedded biometric system, usually fingerprint or facial recognition. Furthermore, in situations requiring more stringent security, biomet- rics are a suitable alternative to protect sensitive data; for example, banks can use them to provide secure access to each account of its clients (Saralaya et al., 2017). There are many types of biological traits used in the development of biometric systems. Nevertheless, many of them are susceptible to brute force attacks, forgery, or direct forcing on users (Prabhakar et al., 2003). Furthermore, a constant problem of most biometrics is that they cannot guarantee that the user is alive (Rui & Yan, 2018). To overcome these problems, the electroencephalogram (EEG) signals are an excellent choice as biometric traits; the main advantages of EEG as a biometric include (Chan et al., 2018): • Universality: All human brains have neurons producing electrical activity that can be read in the form of EEG signals. Individuals ∗ Corresponding author. E-mail addresses: dustin.carrion@yachaytech.edu.ec (D. Carrión-Ojeda), rfonseca@inaoep.mx (R. Fonseca-Delgado), ipineda@yachaytech.edu.ec (I. Pineda). of any age and any mental state, including a vegetative state or coma, produce these signals. • Distinctiveness: The evidence from EEG-based person recognition research shows that EEG signals are sufficiently different among people to classify them. • Permanence: Some session-to-session tests have been conducted to validate EEG variability over time; these studies have con- cluded that EEG signals maintain a significant degree of repeata- bility to develop biometric systems (Maiorana et al., 2016). • Circumvention: One of the most crucial problems of traditional biometrics is that the features can be collected without one’s consent. Nonetheless, no technique allows the recording of brain waves remotely. Additionally, the negative emotions produced by forcing a user to make an EEG recording would lead to an authentication failure (Del Pozo-Banos et al., 2014). Despite the advantages of EEG signals, it is challenging to develop biometric systems based on them. Preprocessing needs to be applied to these signals before working with them because they contain a mixture of frequency bands and artifacts produced by the layers of the human https://doi.org/10.1016/j.eswa.2020.113967 Received 7 July 2020; Received in revised form 1 September 2020; Accepted 2 September 2020