EEG-based person authentication for security systems

  • Tien Dung Pham

    Student thesis: Doctoral Thesis


    Authentication is essential in security operations. There are three popular methods of authentication: authenticating by something a person knows, something a person has, and something a person is or does. Each has its advantages and drawbacks. In this research, we propose to use EEG signals as a new means of authentication which can be known as something a person thinks. This new method is not a replacement for any the popular three authentication methods but has its niche where very high security levels are required. First, an EEG-based person authentication solution for multi-level security systems is proposed where by a person can be verified by different combinations of \EEG passwords" that correspond to level of security. This method improves the performance of the security system overall. It also provides a exible access control policy with the advantages of EEG biometrics which are difficult to fabricate, impossible to be observed or intercepted, unique, un-intrusive, and require live person recording. Second, different from some other biometrics, EEG signals carry rich information about a person such as age and gender. It is proposed to extract this useful information and use it as extra and hidden credentials for person authentication as means of enhancing security systems. Third, a simple to implement and easy to use EEG based person authentication method is proposed. The method takes advantage of EEG artifacts to provide both security and usability for security systems. The final contribution of this thesis is speculating on the stability of an EEG-based person authentication system while EEG signals are known to be sensitive to emotions. Evaluation experiments performed on the benchmark datasets such as Graz 2008, Australian EEG, and DEAP show that the proposed method is promising.
    Date of Award2016
    Original languageEnglish
    SupervisorWanli Ma (Supervisor) & Dat Tran (Supervisor)

    Cite this