Information trustworthiness evaluation based on trust combination

Abstract

Publishing information in a virtual organization (VO) has become too easy due to low barriers; hence development of novel mechanisms to assess the quality of collected information has become a necessity. An evaluator makes such an assessment based on the trust he/she places on the information. This paper presents a model for evaluating information trustworthiness in a data-intensive VO.When some information is derived from various data items gathered from multiple sources (each data item is called an object as used together with the term, subject), it is possible that no data value (called a version of the object) satisfies an evaluator's requirement with regard to information quality, if they are evaluated separately. According to the principle of object trust combination, if the final values of an object calculated by using significantly different methods are similar, then the evaluator places higher level of trust in the results. Intuitively, different versions of the same object that are calculated in different ways but have similar values provides "multiple-proofs" towards their correctness. We assume that a subject has no conflicting information on a given object.This paper uses a formal data structure to represent how a given piece of information (object version) has been formed and develops algorithms (see Section 4) to compare the component structure similarity/dissimilarity between two object versions. This helps in calculating the final trust values of the object.