Google launches new secure multi-party computing open source library to collaborate with data in a privacy-safe manner

Google Inc. continues to invest in new research to drive innovation and protect personal privacy. Earlier this year, the password checker was introduced, a Chrome extension that helps users detect if their username and password entered on the site have been compromised. It relies on an encryption protocol called Privacy Set Intersection (PSI) to match your login credentials to an encrypted database of more than 4 billion credentials, and Google knows that these credentials are not secure. At the same time, it ensures that no one (including Google) knows your actual credentials.

Now Google has launched an open source library for privacy joins and calculations ( Private Join and Compute ) , a new type of secure multiparty computing (MPC) that enhances the core PSI protocol, helping organizations work with confidential data sets while improving privacy. You can check out the project on GitHub:

Collaborate with data in a privacy-safe manner

Many important research, business, and social issues can be derived from a combination of data sets from different parties, each of which has personal information about a set of shared identifiers (such as email addresses), some of which It is common. However, when you are dealing with sensitive data, how can one party get aggregated information about the other party without knowing any personal data of the other party? This is a challenge that Private Join and Compute need to address. Using this encryption protocol, both parties can encrypt their identifiers and associated data and then join them. They can then perform some type of calculation on overlapping data sets to summarize useful information from both data sets. All inputs (identifiers and their associated data) remain fully encrypted and unreadable throughout the process. Neither party disclosed their raw data, but they can still use the calculated output to answer the question at hand. The end result is the only result of decryption and sharing in the form of aggregated statistics. For example, this can be a count, sum or average of the data in the two groups.

Learn more about the technology

Private Join and Compute combine two basic encryption techniques to protect a single piece of data:

Privacy Collection Intersection: Allows both parties to join their collection privately and discover the identifiers they share. We use a variant of the problem of a casual problem, which only marks the encrypted identifier without learning any identifier.

Homomorphic Encryption: Allows certain types of calculations to be performed directly on encrypted data without first having to decrypt it, which preserves the privacy of the original data. Personal identifiers and values ​​are still hidden throughout the process. For example, you can calculate how many identifiers are in a public set, or calculate the sum of the values ​​associated with a tagged encrypted identifier – no need to know anything about the individual.

The combination of these two technologies ensures that only the statistics of the size of the connection set and its associated values ​​(such as the sum) are displayed. Individual items are highly encrypted using a random key and are not provided to the other party or anyone else in their original form.

Use multi-party calculations to solve practical problems

Multi-party computing (MPC) is a long-standing field, but it often faces barriers widely adopted outside of academia. Common challenges include finding effective ways to customize encryption techniques and tools to solve real-world problems.

Google is committed to applying MPC and encryption technology to Google and other more specific real-world issues by providing privacy technology more broadly. We are exploring some of Google’s potential use cases through collaborative machine learning, user security, and aggregated ad measurement.

This is only the beginning of a possible start. This technology can help advance valuable research in a variety of areas where organizations need to work together without revealing any information about the individuals represented in the data. E.g:

  • Public Policy – If the government implements a new health plan in a public school (such as better lunch options and physical education), what are the long-term health outcomes of the affected students?
  • Diversity and Inclusiveness – How does this affect the compensation between demographic companies as the industry develops new programs to bridge the gender and racial pay gap?
  • Healthcare – Does a new preventive drug reduce the incidence of disease when it is given to patients across the country?
  • Car Safety Standards – Does the car manufacturer match the reported reduction in car accidents when adding more advanced safety features to the vehicle?

Private Join and Compute ensure personal information security while allowing organizations to accurately calculate and get useful insights from aggregated statistics. By sharing technology more widely, we hope this extends the use case for secure computing.

Source: Gemi chain