From: Machine learning concepts for correlated Big Data privacy
S. no | Privacy measure | Definition | Limitations |
---|---|---|---|
1 | k-anonymity | This is a mechanism for testing algorithms so that published data restricts what can be disclosed about the properties of entities that are to be secured [1, 2] | Homogeneity-attack, background knowledge [2] |
2 | l-diversity | The equivalence class is also said to l-diversity if the sensitive attribute has at least “l-well-represented” values [2, 3] | Insufficient to prevent attribute disclosure [2] |
3 | t-closeness | The table is said to have t-closeness if the distribution of the entire attribute in the whole table is no more than a threshold t [4] | Complex distribution of a sensitive attribute [4] |
4 | Differential privacy | It is a mechanism to provide an interface between user and database, which protects individual data with the highest mathematical guarantee [5] | In the case of data diversity, it includes too much noise, which reduces the data utility [5] |