Disclosure Limitation of Sensitive Rules

Data products (macrodata or tabular data and microdat a or raw data records), are designed to inform public or bus iness policy, and research or public information. Securi ng these products against unauthorized acces ses has been a long-term goal of the database security research comm unity and the government statistical agencies. Solutions t o this problem require combining several techniques and m echanisms. Recen t advances in data mining and machine l earning algorithms have, however, increased the security r isks one may incur when releasing data for mining from out side parties. Issues related to data mining and security hav e been recognized and investigated only recen tly. T his paper, deals with the problem of limiting disclosure o f sensitive rules. In particular, it is attempted to select ively hide some frequent itemsets from large databases with as little as possible impact on other, non-sensitive frequent i temsets. Frequent itemsets are sets of items that appear in the database “frequently enough” and identifying them is usually the first step toward association/correlation rule or sequential pattern mining. Experimental results are pres ented along with some theoretical issues related to this pr oblem.