Abstract or Summary:
Complex authentication schemes, such as some forms of biometric authentication and context analysis, require large quantities of sensor data and the identify verification can be computationally intensive. This can result in long latencies from the time of the authentication challenge until the authorization decision is determined. This can be worse when there is congestion in the system due to excessive authentication requests, such as at the start of the business day or shift change. Interruptions can impact the user's short term memory, slow down task performance, as well as result in user dissatisfaction with the authentication system. This paper proposes stochastic models to represent the authentication process and offers queue management and load shedding solutions whereby various parts of the authentication process can be addressed to mitigate the delays.