ABSTRACT
In the case of statistical databases, Dorothy Denning and others were the first to study the problem extensively (see [IEEE83]). Here the idea is to give out, say, averages and sums while protecting the individual pieces of data. The census bureau also studied the inference problem in collecting and maintaining census data. However, it was not until the late 1980s
when there were many activities on MLS/DBMSs that the inference problem was investigated extensively. Mor genstern at SRI International [MORG87], Thuraisingham at Honeywell [THUR87], and Hinke at TRW [HINK88] developed the early solutions. Later Thuraisingham at MITRE carried out extensive work on the problem together with Ford, Rubinovitz, and Collins (see [THUR89], [THUR93], and [THUR95]). Since then many efforts have been reported, say, by Marks and Farkas among others (see [MARK96] and [FARK01]). Today the inference problem is resurfacing due to technologies such as data warehousing and data mining. This is because data warehousing and data mining exacerbate the inference problem and they also contribute toward privacy violations via inference. We discuss data warehousing, data mining, and privacy in Part VIII.
