How Multi-Party Computation strengthens the industry’s resilience

24. Januar 2025

The industry is a very complex and also extremely sensitive system. Seemingly small changes may lead to a huge impact elsewhere, as it is already well-understood for the bullwhip effect, for example. Having access to certain information is key in decision-making and often companies would benefit from exchanging data about their business with each other. This usually doesn’t happen mainly for two reasons: First, sharing data might harm the competitiveness of the own business, as potentially sensitive data gets exposed to the public. Second, companies might be accused of cartel agreements, even if done with good intentions.

Privacy enhancing technologies (PETs) overcome these issues by selectively disclosing certain pieces of information. One of these technologies is Multi-Party Computation (MPC) which enables multiple independent servers to jointly compute on private data, yielding more abstract or aggregated results that can be disclosed between participants or even to a broader audience.

Furthermore, MPC incentivises participants to provide real data instead of using inaccurate data, as their data remains private. Compare this setting to other collaborative data processing initiatives where often one central party learns all private information. Participants are well-aware that this one central party will learn their data, so they might decide to not share everything in the centralized setting.

In the SC⁴EU project, MPC is used to realize anonymous surveys. Participants provide their answers on sensitive questions, such as the current demand or whether the inventory is growing or shrinking. Once enough participants have provided answers, the data is aggregated into one big picture of the overall current state of the industry while individual participant’s answers remain private due to MPC.

How does Multi-Party Computation work? The participants answer’s are split into so-called “secret shares” on the participants device before sending the answers. Each individual secret share does not reveal anything about the original answer, but computations can be performed on these secret shares. One secret share is sent to one MPC computation server each and these servers then jointly compute the result that will be revealed to a specific group of people, e.g. the initial survey participants. These results are then further used to compute forecasts to identify potential upcoming crises and ultimately to dampen the bullwhip effect.

This MPC powered survey helps to collect relevant data which would not be possible without the use of MPC due to the reasons mentioned earlier. Overall, this survey system helps to better predict market changes early, allowing the industry to react accordingly and reduce financial losses, therefore strengthening the industry’s resilience.