It might really depend what we are considering "deanonymization". This article is take a hopelessly naive approach to simply see how unique existing data sets are (am I misunderstanding this?), which is not the underlying issue here really.
Let me give a concrete example. Let's say you want to collect application usage information from users of your app. You could a) collect all the information attached to an IP address, and any other nominal information, and store it, or b) you could calculate the valuable usage metrics for the entry, and store it in some demographic bucket, throwing away any other knowledge.
The real problem is that companies want the full data so they can be free to change their models at will. We as users should not expect data to be stored properly, and cryptography tools are our main tool to address this problem.
Let me give a concrete example. Let's say you want to collect application usage information from users of your app. You could a) collect all the information attached to an IP address, and any other nominal information, and store it, or b) you could calculate the valuable usage metrics for the entry, and store it in some demographic bucket, throwing away any other knowledge.
The real problem is that companies want the full data so they can be free to change their models at will. We as users should not expect data to be stored properly, and cryptography tools are our main tool to address this problem.