Behavioural biometric profiling and ambient intelligence

Abstract

In many applications, the most interesting features that we may wish to discover, store and/or exploit in decision making, concerning a human subject, involve the psychology at a variety of levels, of this subject. Parameters, ranging from enjoyment to honesty, important to applications such as adaptive entertainment, through to intrusion detection, are commonly but subtly manifested in an individual's behaviour. As humans, we systematically exploit information gleaned from the observation of others' behaviour, often with astounding levels of success, proving the existence of immensely descriptive data in such observations. Technical attempts to exploit the exact same information have been successful in much more focused and limited applications, demonstrating that at least an initial level of access to this information lies within the capabilities of technology. As a core problem, extraction of useful features from measurements of human behaviour is approached as a pattern recognition problem. Behavioural biometrics may be used for verification, identification or miscellaneous types of classification, without the difference between these applications amounting to a paradigm shift, although the technical difficulty of the problem and the technological sophistication required may change significantly. A critical issue, however, is that such input data cannot, in the general case, be filtered to support one application but not another, which raises even deeper ethical concerns than usual. The total information that can be extracted from behavioural biometric measurements forms an especially rich profile for the subject of the analysis. Since behaviour is easy to observe, many of these techniques are non-intrusive, i.e., the subject may not even be aware of them. This leads us to a major current limitation of practical behavioural biometric systems: their sensing capabilities. This is actually a volatile research issue: from camera and microphone installations to monitoring PC users' input device usage rhythms, better measuring capabilities can lead to much improved behavioural biometric systems. There exists therefore a strong connection with the ambient intelligence vision, which creates the potential for highly advanced applications - but also much more subversive threats. © 2008 Springer Netherlands.