Using Keystroke Analysis as a Mechanism for Subscriber Authentication on Mobile Handsets

The mobile communications industry will experience an evolutionary step within the next two years with the introduction of third generation mobile networks, completing the handset transition from a purely telephony device of the first generation analogue networks into a multimedia multi-purpose mobile communications tool. The ability of these new handsets to store and access sensitive information such as financial records, digital certificates and company records in association with a large handset penetration (864 million subscribers) makes them a desirable target for impostors. The authentication technique for current mobile phones has many weaknesses from a technological and subscriber perspective, and as such non-intrusive and stronger subscriber authentication techniques are required. This study investigates the plausibility of one such technique that of keystroke analysis, comparing and contrasting a number of pattern recognition and neural network based approaches to classification. It was found that neural network-based approaches performed substantially better than the pattern recognition-based approaches with false acceptance and false rejection rates of 3.2%.