DevDiary: Rafael

[Ralfons-06]

No description provided.

[Ralfons-06]

Week 01.05 - 07.05

• Basic research on k-Anonymity and Differential Privacy
• More detailed Research on Approaches for k-Anonymity in the Context of Streaming Data
• Preparation of a short presentation regarding k-Anonymity and Differential Privacy

[Ralfons-06]

Week 08.05 - 14.05

• Deciding on what privacy algorithms to use
• Research for example Implementations of Delta-Doca Algorithm and CASTLEGUARD Algorithm
• Profound research of the CASTLEGUARD Algorithm
• Started writing wiki as knowledge base for the Team

[Ralfons-06]

Week 15.05 - 21.05

• Planning and coordination of CASTLEGUARD Implementation
  • Splitting workload into teams
  • Each Team got a summary of important information and some tasks/requirements
• further work on wiki

[Ralfons-06]

Week 22.05 - 28.05

• research on noising mechanisms
  • exploration of other noising mechanisms that could be beneficial to the differential privacy approach
• wip: perturbation function of CASTLEGUARD

[Ralfons-06]

Week 29.05 - 04.06

• finalize pertubation
• Vacation

[Ralfons-06]

Week 05.05 - 11.06

• Started working on fixing SpliL Bug
• Got in contact with Philip regarding multiple sensitive attributes feature with which was in the end scrapped due to time constraints

[Ralfons-06]

Week of 12.06-18.06

• Still working on fix for SplitL method
• Research to improve merging mechanism used by splitL

[Ralfons-06]

Week of 19.06-25.06

• Fixed SplitL Bug and merged into main branch
• Fixed minor issues with mergeCluster function
• Tested some additional ideas for optimizing the merging mechanism which unfortunately didn't work out.
  • due to time constraints we didn't further approach an optimization for this function

[Ralfons-06]

Week of 26.06-02.07

• More Research on the Domain Bounding Phase by delta which is not to well described in the paper and wasn't implemented in the reference doca implementation we used before.
• Startet implementing the Domain Bounding phase for doca
• Did some research on online quantile estimators and modified a reference implementation of the Greenwald-Khanna quantile estimator (https://aakinshin.net/posts/greenwald-khanna-quantile-estimator/) to fit the purpose of the Domain Bounding phase

[Ralfons-06]

Week of 03.07-09.07

• Implemented the Domain Bounding Phase by delta for Delta Doca
• Fixed minor issues in Doca algorithm

[Ralfons-06]

Week of 10.07-16.07

• Final Presentation (Benchmarks + Slides)
• Added AnonymizationInterface to the CastleGuard implementation
• fixed minor bugs
• Add switch for Domain Bounding On/Off

[Ralfons-06]

Week of 17.07-23.07

• prepared for final presentation
  • benchmarks, plotting and slides

[Ralfons-06]

Vorlesungsfreie Zeit

• Vacation: 05.08 - 13.08
• Refactored Pipe class
• Refactored Doca Algorithm to an OOP structure (instead of the functional approach of our previous implementation)
• While doing this we also decided to restore the old approach since the redis-cache approach could be interesting for future work
• Added field prioritization to both algorithms (based on this work: https://ieeexplore.ieee.org/abstract/document/9340178)
• Introduced Abstract Classes for Items and Clusters for better structure and simplified integration of new anonymization algorithms