infonet-economy

Emerging computing infrastructures are heterogeneous, ubiquitous and mobile. Devices from personal computers, to handhelds, to printers, to embedded devices are very widely available. Further, today's wireless network infrastructures make it possible for devices to spontaneously interact. Many devices are mobile, carried by people or mobile machines. Thus, the environment available to a device is constantly changing. This notion also encapsulates issues such as security policy and physical properties such as network signal strength, or printers availability. By their heterogeneity, scale, and dynamism, these systems gain to be designed so that they organise themselves autonomously. Usually interaction among devices that do not know each other occurs through some exchange of information specifying their respective capabilities. However, in a dynamic and unsecure environment, this is not sufficient. On the one hand, a malicious entity can exhibit desirable characteristics, while it is not able to realise them. On the other hand, even if in good faith, a printer can fail because it lacks toner or paper. The approach proposed in this paper is to combine information that entities carry about themselves, with trust or reference information exchanged among entities about other entities. Trust enables adaptation of running entities to the dynamic modifications of their environment. For instance, in the case of users and printers, a printer exhibits its characteristics, such as postscript, double-sided, black and white, and users exchange trust information about printers based on their observations and experiences realised with the printers, such as frequent paper jams, or low toner. Trust circulates among users about printers. Trust in a well functioning printer raises, while trust in an always broken down printer decreases. This paper shows first that the human notion of trust is actually a self-organising mechanism according to Nobel Prize Prigogine criteria for self-organisation. Second, it formally proposes an interaction mechanism based on a tag-based model where entities are equipped with a marking, carrying a specification of the functional as well as non-functional capabilities they offer to the community; and on a trust-based model, where trust information are exchanged among principals. The paper demonstrates as well the approach through a small example involving users and printers. Finally, the paper discusses notions inherent to the human notion of trust, such as belief and confidence in entities expressing trust, as well as the notion of risk which is intimately linked to the notion of trust.