research

Enigma: data sharing without liability

research

Enigma: data sharing without liability

April 2020

Posted by

Theo Turner

Technology Analyst

Theo is an R&D software developer working with blockchain, machine learning and the internet of things. In addition to building applications for internal use, Theo completes technical due diligence on new projects and assists Outlier’s portfolio companies with technical capabilities....read more

In 2020’s economy, business decisions are driven by data. Demand for data is higher than ever, with organizations increasingly relying on external data sources and data sharing to capture value. However, in the GDPR era, data liability is a substantial risk, leading to a wealth of missed opportunities in data due to privacy concerns. Read on to learn about our continued work with Enigma, who are rethinking data security to encourage data sharing while eliminating data liability risk.

Enigma is developing a blockchain-based protocol for sharing data responsibly by preserving privacy not only during transmission but also at compute time. We’re excited to work with Enigma, as they are leveraging Trusted Execution Environments (TEEs) that allow organisations to share data without risk of it being revealed, even to the parties they choose to share their data with. With Enigma, enterprises can gain insights about data shared between their partners without direct access to any of the data.

In a previous blog post, we detailed a collaborative analytics product we have been building on Enigma, which we have dubbed APEX (Analytics from Private Execution). Machine learning, which is increasingly helping us to make better business decisions, is the core focus of the software. The added ability to share data without liability allows us to further improve these business decisions, and we’re now opening up what we’ve built to the world. To expand on our previous post, we’ll cover the data privacy aspects of the software, enterprise use cases, as well as updates to the software, including a slick new UI.

The Enigma network enables cryptographic guarantees of privacy by encrypting data before it leaves enterprise premises. When inputting any data to the network, it is first encrypted client-side, so we can be confident that there is no way for the data to be read in transport. The data is decrypted in a secret contract, a piece of publicly verifiable code at a public, immutable address on a blockchain. Enterprises can be entirely confident that their data is not able to be viewed by anyone, or that it can be intercepted by man-in-the middle attacks.

Multiple enterprises can send their data into the same secret contract to share data without revealing it. The contract, with publicly verifiable compute logic, can output results based on the sum of all shared data without revealing any individual particpant’s dataset. With APEX, we have written secret contracts which enable enterprises to perform machine learning on shared data. We’re opening up the tooling that has helped us make better business decisions to the wider corporate space, and supercharging it by allowing computation on shared datasets. All of this without data liability risk, thanks to Enigma.

APEX allows enterprises to perform two different forms of machine learning on shared data, as well as collaborative model training. The first form of machine learning we chose to include is clustering, a way of grouping data according to their closeness to relative density centers. A simple example are the inhabitants of various cities in a country, which may be clustered into, or grouped by, their respective cities. Each cluster comes with a centroid, a relative center of density, which in our example corresponds to the center of each city. Clustering is a powerful form of machine learning for grouping a variety of data types, but is particularly potent when it comes to geographic customer data. Our analytics platform allows clustering to be performed on latitude/longitude data, with the flagship use case for telecommunications infrastructure placement: by clustering customer locations, telco providers learn the optimal placements of cell phone towers to guarantee coverage across their networks, or optimal retail store locations to be as close to as many customers as possible. With multiple enterprises sharing data, businesses get analytics for customers they don’t even know about, all without those customers being revealed to them, thanks to Enigma. The clustering in APEX can be generalized to a range of other geographic optimisation challenges, such as car routing for ride sharing applications, and even non-geographic problems such as identifying fraudulent transactions.

The second form of machine learning available in APEX is classification. Classification allows for concrete segmentation of data points into classes, as the name implies. Unlike clustering, classification does not judge by relative density of data, but by training data input by its users. In the case of our software, that training data may be contributed by enterprises, allowing them to classify customers according to any metric of their choice, such as jurisdiction. A particularly strong point of Enigma’s data sharing without liability is the possibility to train machine learning models collaboratively. This allows enterprises to create more powerful AI than what has been possible previously, as the AI is able to learn from the sum of all knowledge of the enterprises contributing their data. This allows enterprises to gain insights about all of the potential customers in an industry, rather than just their own customers. For example, our software allows telco providers to know the total size of the addressable market in different countries far more accurately than before, as the result is based on real data, rather than estimations. Collaborative model training can be further extended to a myriad of use cases for machine learning, such as crop yield predictions and insurance fraud identification, featuring significantly improved accuracy and quality of insights thanks to Enigma data sharing.

Sign up for the launch event of APEX here.

Posted by Theo Turner - April 2020

April 2020

Posted by

Theo Turner

Technology Analyst

Theo is an R&D software developer working with blockchain, machine learning and the internet of things. In addition to building applications for internal use, Theo completes technical due diligence on new projects and assists Outlier’s portfolio companies with technical capabilities....read more