Independence in information spaces (Q1928485)

The paper discusses dependencies in information spaces. An information space is any non-empty set the elements of which represent pieces of information, or ``secrets'' according to the terminology used in the paper. For example, an information space can be associated with a relational database, or with the information exchange of two or more intelligent agents. Understanding dependencies between secrets is thus important in several respects. Concerning databases, users are usually associated with specific privileges and are often allowed to query only specific tables. Concerning information exchange, agents might be interested in preserving sensitive secrets. In particular, the paper introduces a new type of interdependence satisfying the following properties: {\parindent=0.5cm\begin{itemize}\item[--] any set of secrets is independent from the empty set; \item[--] independence is a symmetric relation; \item[--] if two sets \(A,B\) of secretes are independent, then \(A\) is independent from any subset of \(B\); \item[--] public knowledge, i.e., attributes associated with constant functions, does not introduce dependencies. \end{itemize}} A complete and decidable axiomatization of this new independence relation is provided.