Design and Deployment of Identity Recognition Systems

Modern Identity Recognition Systems (IRSs) can benefit from the simultaneous use of different biometric parameters but their activity cannot be modeled using a rigid sequential path because they have to take into account outcomes from different data as well as from different algorithms at different times. This paper is about the representation of an ad hoc composition of elementary recognition service in order to form an integrated recognition facilities, to be deployed within the scope of energy-aware systems. We aim to investigate the types of useful collaboration among IRSs components and to present a tool to assist designers both in the design process and in the system deployment phase. The tool define a catalogue of elementary recognition services that are typically associated with single trait recognition modules. They are considered as prototypical services that can be differentiated by proper parameter settings. Those basic building blocks can form compound collaborative patterns: composition rules are expressed both as logical association and causal relation (what-if rules). The recognition activity can be organized through different levels and feedback is allowed. Partition let to evaluate the overall QoS from performance figures that are associated with the elementary blocks, in order to certify expected Service Level Agreements (SLA)s. Deployment represents an important step before system release. The proposed approach is an interactive one: system administrators are asked to allocate all the components on actual machines and secure communication channels are automatically established. Scalability, redundancy and fault tolerance and energy issues can be addressed at this level.