In real life situations data is kept in multiple relational tables instead of single table. In relational tables, not all sensitive or identifying information is kept in each table separately. In this case k-anonymity may not be the best option. It may fail to ensure anonymity or cause too much distortion making data useless for data miners. Instead of running k-anonymity on each table separately, in Multirelational k-Anonymity tables are first joined together using some identifier that is common for each table.
For example suppose we have person specific table called Person, table called Diasese that contains list of diaseses and table called Job that lists jobs where people work. In these tables Id field is common for all three tables.
| Id | First name | Last Name | Age | Gender | City |
|---|---|---|---|---|---|
| 1 | Juhan | Olev | 33 | Male | Elva |
| 2 | John | Smith | 36 | Male | Elva |
| 3 | Gebede | Ahmed | 40 | Male | Tartu |
| Id | Diasese |
|---|---|
| 1 | Flu |
| 2 | Hiv |
| 3 | Aids |
| Id | Job |
|---|---|
| 1 | Programmer |
| 2 | Analyst |
| 3 | Developer |
In this example we can join tables using two rules: Person.Id = Diasese.id and Person.Id
= Job.Id. These rule
would give us the following table:
| Id | First name | Last Name | Age | Gender | City | Diasese | Job |
|---|---|---|---|---|---|---|---|
| 1 | Juhan | Olev | 33 | Male | Elva | Flu | Programmer |
| 2 | John | Smith | 36 | Male | Elva | Hiv | Analyst |
| 3 | Gebede | Ahmed | 40 | Male | Tartu | Aids | Developer |
At left you have tables that you selected in previous step. Your task is to specify rules for joining these tables. To do this these steps:
Add rule. Now you should see that table headers have turned from gray to
black an you can click on them.
Id column on
Person table in previous step you might now want to click on Id column on
Diasese table.
Ok to finish defining of rule after which you can see rule in purple box,
Below the tables there is section called Preview of join which you can use to check if you have
deined rules correctly. This table will be updated as you add or remove rules.
If you wish to remove the rule simply click on red X in rule box.
Once you have defined all the necessary rules, click on next button.