Trust is a property that can depend on a range of other properties. For example, trust can be influenced by the perceived dependability of the system, its security measures, or the behaviour of other users of the system. Actually defining and measuring trust is a difficult task due to its highly subjective nature. For example, a pilot might ‘trust’ his co-pilot in every day life, but when preparing to fly a plane he has to intentionally not trust him and double check all his pre-flight checks. Furthermore, attempting to quantitatively measure this level of trust is also practically (if not totally) impossible. Because of this no attempt is made to define it here.
In traditional computing trust is implemented through certification authorities, entities that can vouch for a certain user/device. Certificate authorities (CA) are linked together in a hierarchal fashion to form ‘chains of trust’ to allow users/devices to authenticate and administer a domain of trust between themselves and others within that chain.
The structures of these ‘chains of trust’ are built upon the current commercial Internet infrastructure. P2P systems are essentially ad hoc adaptive networks, and using centralised certificate authorities for authentication purposes is quite obviously a major problem, as it would involve knowledge of the network infrastructure in advance.
To allow the inclusion of CA’s within a P2P infrastructure would likely involve the leverage of some aspect of the network architecture and/or application domain.