Making agents secure on the semantic Web

Agents were designed to collaborate and share information. While highly desirable for interoperability, this feature is scary from the security perspective. Illegal inferences, supported by semantic Web technology and ontologies, might enable users to access unauthorized information. In addition to semantic associations and replicated data with different sensitivity, malicious agents could also exploit statistical inferences. Although each agent in a system might behave in a desired and secure way, their combined knowledge could be used to disclose sensitive data. The research community must therefore develop and implement techniques that allow control over released data. To answer the questions related to information availability (scalability), data correctness (integrity), and access control in the presence of illegal inferences and undesired collaborations (confidentiality), researchers in semantic Web technologies (XML, RDF, DAML, and multiagent systems) and information system security need to collaborate. Indeed, given the Web´s openness, dynamic nature, and diverse user population, developing secure Web services will require the collaboration of experts in different fields from both industry and academia. In turn, the intelligent Web of the future will facilitate unheard of support for collaborations and information management.