Establishing Trust Beliefs Based on a Uniform Disposition to Trust

Abstract

Abstract

Composing Trust Models towards Interoperable Trust Management

“…The concatenation of 1:1 trust relations usually represents a transitive trust path, where each entity can trust unknown entities based on the recommendation of its trustees. Thus, this relationship is built by composing personal trust relations [1,18]. Furthermore, in the case where there exist several trust paths that link the trustor to the recommended trustee, the aggregation can be used to aggregate all given trust recommendations [7].…”

“…Other solutions represent trust by a numerical range. For instance, this range can be defined by the interval [-1..1] (e.g., [12]), [0..n] (e.g., [1,18]) or [0..1] (e.g., [7]). A trust value can also be described in many dimensions, such as: (Belief, Disbelief, Uncertainty) [7].…”

“…Most existing solutions simply initialize trust relation with a fixed value (e.g., 0.5 [6], a uniform Beta probabilistic distribution [8]). Other approaches include among others: initializing existing trust relations according to given peers recommendations [17]; applying a sorting mechanism instead of assigning fixed values [18]; and assessing trustees into different contexts (e.g., fixing a car, babysitting, etc.) and then inferring unknown trust values from known ones of similar or correlate contexts [16,2].…”

“…All the solutions dealing with 1:N trust assessment mainly define the concatenation and the aggregation operations, in order to concatenate and to aggregate trust recommendations by computing the average [18], the minimum or the product [1] of all the intermediary trust values. In the case of Web service composition, some approaches (e.g., [15]) evaluate the recommendation for each service by evaluating its provider, whereas other approaches (e.g., [11]) evaluate the service itself in terms of its previous invocations, performance, reliability, etc.…”

“…In the algorithm, the mediator interacts with r s (see lines 2-4) and r t (see lines 5-7). Then, the mediator computes the matching relation for each output relation (see lines [11][12][13][14][15][16][17][18] of the reply, where we assume that there is only one such relation (see lines 12-13) and requests its value using the appropriate request operation (see lines [16][17][18]. We further consider that the mediator (µr) embeds a library of mediation functions that translate and normalize heterogeneous trust metrics, which are invoked by mediation operations µo (see lines [12][13][14].…”