[MADLIB-911] Anonymization

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Module: Utilities
• Labels: starter
• Assignee: Unassigned
• Reporter: Frank McQuillan
• Created: 02/Nov/15 22:40
• Updated: 13/Nov/18 18:03

Description

Story

As a data scientist, I want to perform anonymization operations on my data, so that I can prepare it for input to predictive analytics algorithms. I also want to be able to de-anonymize my data.

This feature is relevant especially given the recent GDPR policy:
https://eugdpr.org/

Proposed functionality:

• Create conversion table for anonymization.
• Create an anonymized version of a table.
• Create a deanonymized version of a table

Must be able to:

• anonymize multiple columns in a table
• datasets will still join correctly even on masked columns
• the aggregates on masked columns will match to the original
• add salt to hash function for better security

References

[1] PDL tools
http://pivotalsoftware.github.io/PDLTools/group__grp__anonymization.html

[2] General information on anonymization
https://en.wikipedia.org/wiki/Data_anonymization

[3] Blog on hashing
https://crackstation.net/hashing-security.htm

Issue Links

• mentioned in: Page Loading...
• mentioned in: Page Loading...