Simultaneous Pattern and Data Hiding in Unsupervised Learning

How to control the level of knowledge disclosure and se- cure certain confidential patterns is a subtask comparable to confidential data hiding in privacy preserving data min- ing. We propose a technique to simultaneously hide data values and confidential patterns without undesirable side effects on distorting nonconfidential patterns. We use non- negative matrix factorization technique to distort the origi- nal dataset and preserve its overall characteristics. A fac- tor swapping method is designed to hide particular confi- dential patterns for k-means clustering. The effectiveness of this novel hiding technique is examined on a benchmark dataset. Experimental results indicate that our technique can produce a single modified dataset to achieve both pat- tern and data value hiding. Under certain constraints on the nonnegative matrix factorization iterations, an optimal solution can be computed in which the user-specified con- fidential memberships or relationships are hidden without undesirable alterations on nonconfidential patterns.