Secure XML Publishing without Information Leakage in the Presence of Data Inference

Recent applications are seeing an increasing need that publishing XML documents should meet precise security requirements. In this paper, we consider data-publishing applications where the publisher specifies what information is sensitive and should be protected. We show that if a partial document is published carelessly, users can use common knowledge (e.g., "all patients in the same ward have the same disease") to infer more data, which can cause leakage of sensitive information. The goal is to protect such information in the presence of data inference with common knowledge. We consider common knowledge represented as semantic XML constraints. We formulate the process how users can infer data using three types of common XML constraints. Interestingly, no matter what sequences users follow to infer data, there is a unique, maximal document that contains all possible inferred documents. We develop algorithms for finding a partial document of a given XML document without causing information leakage, while allowing publishing as much data as possible. Our experiments on real data sets show that effect of inference on data security, and how the proposed techniques can prevent such leakage from happening.