Nathan Clarke



Recent years have witnessed widespread adoption of mobile devices. Whereas initial popularity was driven by voice telephony services, capabilities are now broadening to allow an increasing range of data orientated services. Such services serve to extend the range of sensitive data accessible through such devices and will in turn increase the requirement for reliable authentication of users. This thesis considers the authentication requirements of mobile devices and proposes novel mechanisms to improve upon the current state of the art. The investigation begins with an examination of existing authentication techniques, and illustrates a wide range of drawbacks. A survey of end-users reveals that current methods are frequently misused and considered inconvenient, and that enhanced methods of security are consequently required. To this end, biometric approaches are identified as a potential means of overcoming the perceived constraints, offering an opportunity for security to be maintained beyond pointof- entry, in a continuous and transparent fashion. The research considers the applicability of different biometric approaches for mobile device implementation, and identifies keystroke analysis as a technique that can offer significant potential within mobile telephony. Experimental evaluations reveal the potential of the technique when applied to a Personal Identification Number (PIN), telephone number and text message, with best case equal error rates (EER) of 9%, 8% and 18% respectively. In spite of the success of keystroke analysis for many users, the results demonstrate the technique is not uniformly successful across the whole of a given population. Further investigation suggests that the same will be true for other biometrics, and therefore that no single authentication technique could be relied upon to account for all the users in all interaction scenarios. As such, a novel authentication architecture is specified, which is capable of utilising the particular hardware configurations and computational capabilities of devices to provide a robust, modular and composite authentication mechanism. The approach, known as IAMS (Intelligent Authentication Management System), is capable of utilising a broad range of biometric and secret knowledge based approaches to provide a continuous confidence measure in the identity of the user. With a high confidence, users are given immediate access to sensitive services and information, whereas with lower levels of confidence, restrictions can be placed upon access to sensitive services, until subsequent reassurance of a user's identity. The novel architecture is validated through a proof-of-concept prototype. A series of test scenarios are used to illustrate how IAMS would behave, given authorised and impostor authentication attempts. The results support the use of a composite authentication approach to enable the non-intrusive authentication of users on mobile devices.

Document Type


Publication Date