Use of random time-intervals (RTIs) generation for biometric verification

• Published in: Pattern recognition 2009-11, Vol.42 (11), p.2787-2796
• Main Authors: Laskaris, Nikolaos A., Zafeiriou, Stefanos P., Garefa, Lambrini
• Format: Article
• Language: English
• Subjects: Applied sciences; Authentication; Cognitive biometrics; Cryptography; Exact sciences and technology; Information, signal and communications theory; Multivariate Wald-Wolfowitz test; Pattern recognition; Reconstructed dynamics; Signal and communications theory; Signal processing; Signal representation. Spectral analysis; Signal, noise; Telecommunications and information theory
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2008.12.028

We explore the possibility of using human-generated time-series as biometric signature. Adopting a simple psychometric procedure, in which a button is pressed in entirely random manner, successive elapsed times are registered and gathered in a signal reflecting user's internal cognitive processes. By reconstructing and comparing the dynamics across repetitions from the same subject a noticeable consistency was observed. Moreover, the dynamics showed a prominent idiosyncratic character when realizations from different subjects were contrasted. We established an appropriate similarity measure to systematize such comparisons and experimentally verified that it is feasible to restore someone's identity from RTI (random time-interval) signals. By incorporating it in an SVM-based verification system, which was trained and tested using a medium sized dataset (from 40 persons), a considerably low equal error rate (EER) of ∼5% was achieved. RTI signals can be collected effortlessly and this makes our approach appealing, especially in transactions mediated by standard pc terminal keyboards or even telephone keypads.