A New K-anonymity Algorithm towards Multiple Sensitive Attributes

While k-anonymity algorithm has been used widely to prevent privacy disclosure in datasets published with single sensitive attribute. We improve k-anonymity algorithm to protect privacy in multiple sensitive attributes. Based on greedy strategy, we sort sensitive attributes and tuples first. We distribute tuples to equivalence classes evenly according to sensitive values in high degree. We use statistic information to cut off association among sensitive attributes to prevent positive and negative privacy disclosure. Information entropy is introduced as metric of diversity. Experiments on real dataset showed that our algorithm is effective and efficient.