Is 2020 finally the year of the PETs (Privacy Enhancing Technologies)?


Sharing is caring, even for personal data

Multi-party computation might make it easier to share personal data

To maximize the potential of data, organizations typically want to share and combine data sets. However, privacy concerns can pose problems when doing so. Read our blog how to safely share and analyze personal data, without invading privacy.

Written by Zhasmina Kostadinova

The challenge

Here’s a problem many organisations face these days: they want to share personal data between different departments within their organisation, or sometimes even share personal data with other organisations, to find solutions to common challenges. However, sharing personal data can be problematic from a privacy perspective: personal data may end up in places or be used for purposes in a way that it is not transparent or fair towards the individual (and it may not be lawful). 

This is why many organisations leave that part of the potential of their data unused. Secure multi-party computation (MPC or SMC, hereafter MPC) allows you to access that potential. In this second blog of our series on PETs (Privacy Enhancing Technologies) we explain how MPC can support your organisation’s needs.

So, what is MPC?

To explain what MPC does, we introduce to you the Millionaire’s Problem: two millionaires want to know which of them is richer without revealing their net worth to each other. Classically, you would say: “Ok, let’s find a (trusted) third party, have both millionaires submit their net worth and the third party will declare which of them is richer.” 

That works, but for the fact that the millionaires have lost some of their privacy. A third party now knows their net worth. Irrespective of how trusted that third party may be, a man-in-the middle might not be desirable. 

This is where MPC can offer a solution. It can do the exact same thing but without the third party and with the crucial bit of information (exact net worth) not being handed over by the millionaires to anyone else. All it takes is for both millionaires to run a MPC protocol in conjunction and it will spit out the result: “smaller than”, “exactly equal to”, or “larger than”. Information is still exchanged between the two millionaires when they run the protocol, but none of it will reveal anything about their net worth – and that can be mathematically proven (see here). This way, the millionaires have precisely the information they want, without sharing the information they want to keep to themselves.

The strong point of MPC is that it can be expanded to more than two parties and to other questions than comparing numbers. With that in mind, MPC can tackle the privacy issues of many cases of data sharing and analytics, as the personal data will be no longer exchanged or brought together to come to a certain conclusion. 

How can MCP support your business?

The key idea of MPC was introduced in the early 1980s, however only over the last decade have algorithms and computing power become fast enough to actually deploy it. Today, MPC has found many real-world applications. Some of the Privacy Enhancing Technology applications of MPC are:

  • Private set intersection, where two or more companies can determine if some user occurs in both user-databases without processing that user’s personal data without their consent. 
  • Private statistics, i.e. computing statistics via MPC on personal data across various data sources (e.g. different organizations such as hospitals, banks or insurance companies).  
  • In 2015, US researchers used MPC software to analyse the gender wage gap in Boston by securely collecting and analysing payroll data, revealing that the city’s women make 77 cents for every dollar a man makes.
  •  Privacy-preserving machine learning applications, such as secretly analyzing the difference between the input and output of a neural network in order to evaluate its performance. 
  • Enabling technology for threshold cryptography such as storing secrets (e.g. cryptographic keys) in different locations to make accessing valuable information very difficult. Moreover, if you are a multinational company that has to adhere to different legislation in various jurisdictions, MPC can offer secure sharing of data and gathering of information without compromising both customers and employees’ data according to the rules of each jurisdiction. You can find more information on the applications of MPC here.

MCP as a fit for your business

Like any existing tool, MPC comes with benefits and drawbacks. The key to successfully leveraging its capabilities is spotting the right opportunity for implementation. The following general statements could help you to decide whether you want to hear more about MPC.


MPC provides very strong input privacy, meaning that the personal data is not shown to other parties. On top of that it gives correctness guarantees, meaning the output is correctly computed from the given inputs. These properties are ensured in theory by mathematical proofs, and can be assured in practice by a software audit.


Computation in MPC is typically more time-consuming than computation on a single source. The performance depends (among other factors) on the nature of the computation and on the latency and throughput of the computer network that interconnects the parties. Despite increases in algorithmic efficiency, performing MPC on a set of data can still be tens of times slower in comparison to an operation on unprotected data in a single source. 

Notwithstanding the pros and cons that we mentioned, the potential applications of secure multi-party computation are incredibly diverse. Currently, we see viable applications in the financial and insurance sectors with a risk management focus, and in the marketing research sector for unlocking bigger data sets. But also beyond these examples we can quickly assess the practicality of MPC against your business case.

Did you find this useful?