MLforDifferentialPrivacy

• 

  Heat Map of Correlation in the newly generated data, which closely matches that of the original data.

Inspiration

As a team without much experience in Differential Privacy, we wanted to explore the techniques which protect the privacy of data, especially in the age of Big Data.

What it does

Our model processes the yellow cabs data and transforms certain critical fields which have consequences on privacy while preserving distributional statistics of the original data. The model observes correlations between various attributes and transforms attributes that are correlated in a similar fashion. We use Lagrangian noising for continuous attributes and multi-class classification for categorical attributes.

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for MLforDifferentialPrivacy

Built With

• numpy
• pandas
• python
• sklearn