Publications

David Walling, TechNewsWorld, 12/21/07

Data mining isn't always about structured data. Text mining -- or text data mining -- is about comprehending natural language and extracting high quality information from it. Natural languages have structure, too. These structures are generally more complex than a schema, especially one designed for data mining.

entire technologies have arisen to extract, parse and analyze text. At the same time, an increasing amount of stored data is becoming subject to privacy measures, especially encryption.

Clearly, data mining operations must access the plain expression of meaning, or plaintext, in order to mine it for useful information. In the case of structured data, the unit subject to encryption may be a relatively small set of symbols, perhaps only a field or row. When data is structured, encryption can be efficiently applied in various dimensions.

... Approaches to Secure Mining

It may seem strange to contemplate allowing encrypted text to be mined at all. However, text that is valuable for mining isn't necessarily public information.

Furthermore, mining text may not necessarily compromise data security, considering that the result of data mining may simply be an aggregation -- or the rules that govern an inference engine or a neural network -- rather than the details of the text itself. Consequently, some form of control is necessary because data is not always mined by its owner.

Read more.