(δ,l)-diversity: Privacy Preservation for Publication Numerical Sensitive Data

(ε,m)-anonymity considers ε as the interval to define similarity between two values, and m as the level of privacy protection. For example {40,60} satisfies (ε,m)-anonymity but {40,50,60} doesn't, for ε=15 and m=2. We show that protection in {40,50,60} sensitive values of an equivalence class is not less (if don't say more) than {40,60}. Therefore, although (ε,m)anonymity has well studied publication of numerical sensitive values, it fails to address proximity in the right way. Accordingly, we introduce a revised principle which solve this problem by introducing (δ,l)-diversity principle. Surprisingly, in contrast with (ε,m)-anonymity, the proposed principle respects monotonicity property which makes it adoptable to be exploited in other anonymity principles.