文摘
To protect an agent’s own knowledge or belief against unwanted information inferences by cooperating agents, Controlled Interaction Execution offers a variety of control methods to confine the information content of outgoing interaction data according to agent-specific confidentiality policies, assumptions and reaction specifications. Based on preliminary experiences with a prototype implementation as a frontend to a relational DBMS, in this article we outline the architectural design and the parameterized construction of specific tasks to uniformly shield all information services in need of confinement, potentially comprising query answering, update processing with refreshments, belief revision, data publishing and data mining. Refraining from any intervention at the cooperating agents, which are also seen as intelligently attacking the defending agent’s own interest in preserving confidentiality, the engineering solely aims at self-confinement when releasing information.