Language-independent gender identification through keystroke analysis. (13th July 2015)

• Record Type: Journal Article
• Authors: Tsimperidis, Ioannis; Katos, Vasilios; Clarke, Nathan
• Is Part Of: Information and computer security. Volume 23:Number 3(2015)
• Page Start: 286
• Page End: 301
• Publication Date: 2015-07-13
• Subjects: SVM -- Bayesian -- Gender recognition -- Keystroke dynamics -- Manhattan classifier; Computer security -- Management -- Periodicals; Computer networks -- Security measures -- Periodicals; Data protection -- Management -- Periodicals; 658.47
• DOI: 10.1108/ICS-05-2014-0032
• Languages: English
• ISSNs: 2056-4961

Abstract

Abstract : Purpose: – The purpose of this paper is to investigate the feasibility of identifying the gender of an author by measuring the keystroke duration when typing a message. Design/methodology/approach: – Three classifiers were constructed and tested. The authors empirically evaluated the effectiveness of the classifiers by using empirical data. The authors used primary data as well as a publicly available dataset containing keystrokes from a different language to validate the language independence assumption. Findings: – The results of this paper indicate that it is possible to identify the gender of an author by analyzing keystroke durations with a probability of success in the region of 70 per cent. Research limitations/implications: – The proposed approach was validated with a limited number of participants and languages, yet the statistical tests show the significance of the results. However, this approach will be further tested with other languages. Practical implications: – Having the ability to identify the gender of an author of a certain piece of text has value in digital forensics, as the proposed method will be a source of circumstantial evidence for "putting fingers on keyboard" and for arbitrating cases where the true origin of a message needs to be identified. Social implications: – If the proposed method is included as part of a text-composing system (such as e-mail, and instant messaging applications), it could increase trust toward the applications that use it and may also work as a deterrent for crimes involving forgery. Originality/value: – The proposed approach combines and adapts techniques from the domains of biometric authentication and data classification.