Identification and Retrieval of Personal Records from a Statistical Data Bank

From a statistical data bank containing anonymous but individual records which may be evaluated by dialogue—e.g., by entering logical conditions which the system answers by returning absolute frequencies—personal records may often be retrieved. First a user has to identify the record in question, using previous knowledge. An experiment with authentic data is described providing some information on the amount of preknowledge needed for identification. Once identified, parts or the whole of a record may be retrieved by various techniques. The efficiency of a technique already known is studied. A new technique permitting retrieval of complete personal records is described, which is resistant to conventional output restrictions. It becomes feasible as soon as AND plus at least one of the logical operations NOT and OR are available. Some possible countermeasures are discussed.