Google project lets companies collaborate privately in encrypted datasets

Google has set up a project that allows multiple companies to collaborate in datasets without disclosing information to each other. Private Join and Compute uses homomorphic encryption, which allows users to compare encrypted results.

The project’s source code is publicly available on GitHub. Private Join and Compute is mainly intended for researchers, but companies can also get started with the technology, writes Google in a blog post. The company has also released a paper detailing the research.

For the project, Google uses a combination of private set interaction, with which users can share data sets among themselves. Private Join and Compute also uses homomorphic encryption, an encryption standard that dates back to the 1970s. Homomorphic encryption allows the software to perform calculations on an encrypted text in certain cases. The result itself is also encrypted, but users can compare it with the results of other calculations, also on other data sets. That way, data can remain secret even if someone has the encryption key.

As an example of what the project is for, Google mentions a shopping area where the municipality wants to know whether it is worth running an extra train service there. To find out whether the costs of that train service outweigh the revenues of retailers, both parties must have each other’s data. Google’s method ensures that calculations are possible on both data sets without the parties actually seeing the data. The encryption method has been around for decades, but according to Google, it was far from practical and mainly theoretical.

Google has also been working for some time to analyze personal data of users without seeing that data itself. The company used a similar technology for the Password Checkup, an extension that allows users to see if their password has been found in a data breach. The company also wants to apply this to search results in the future.