tf2-published-models VS differential-privacy

- https://github.com/sarus-tech/tf2-published-models

We plan to continue building trust in the tools we are using by publishing some of them.

An example is probably easier :) I quote here the description of the Google's differential privacy example:

GitHub: https://github.com/google/differential-privacy/tree/main/learning/clustering

Maybe it is not so radical. The original, pre-web internet was not client-server. Each end of the connection potentially had something the other wanted. IMO, that's a truer representation of the real world. Today's internet is entirely web and mobile app centric, as if the world is nothing more than a feedlot, with only a small number of large scale "farmers".

https://github.com/google/differential-privacy/blob/main/exa...

One of the most durable techniques to protect user privacy is through differential privacy. In a previous post, we explored how to build an Apache Beam pipeline that extracted and counted ngrams from HackerNews comments. Today, we'll take the same pipeline and upgrade it with some differential privacy goodness using Privacy-on-Beam from Google's differential privacy library.