How to encrypt specific attributes

Satoshi_Kitajim

I have a question about data encryption. We analyze personal information data, but we would like to encrypt all ID attributes. If you only know the ID, you can link it to various data. High security environment is required in our project.

We would like to encrypt the ID key once, then share to some analyst, and only the administrator can decrypt it.

Do you have those features?

I found the following extension but it doesn't seem to be supported in the latest version.

Cryptography Extension
https://marketplace.rapidminer.com/UpdateServer/faces/product_details.xhtml?productId=rmx_cryptography

Best regards
Satoshi

BalazsBarany

Hi,

without the cryptography extension you could use the following approach:

- Generate an additional ID column for your data using the Generate Id operator. Check if there is a correlation between the generated and the original IDs. If yes, shuffle the data first before generating the ID.

- Save the example set with the original IDs and the generated IDs for your administrator (e. g. in a CSV file).

- Give the data to the analyst without original IDs

- Join back the results with the original ID.

Regards,

Balázs

BalazsBarany

Hi,

without the cryptography extension you could use the following approach:

- Generate an additional ID column for your data using the Generate Id operator. Check if there is a correlation between the generated and the original IDs. If yes, shuffle the data first before generating the ID.

- Save the example set with the original IDs and the generated IDs for your administrator (e. g. in a CSV file).

- Give the data to the analyst without original IDs

- Join back the results with the original ID.

Regards,

Balázs

Satoshi_Kitajim

Hi Balázs,

Thank you very much for your answer.  I was looking for operators and extensions.  I think that the problem can be solved by giving the table encryption and decryption information.

Marco_Boeck

Hi,

Attaching any sort of encryption information to data objects defeats the purpose. You need to have the encrypted data and the key separately, otherwise the entire exercise is just security by obscurity (aka no security).

We do not offer support for storing encrypted data tables at this point and likely never will, as that makes collaboration very difficult. There are more effecive means to achieve data security, e.g. by only storing data in DBs with strict access controls.

Balazs outlined an effective and secure way of handling your scenario, given that the file linking both together is kept secure.

Regards,
Marco