Pillar VC
Published in

Pillar VC

Machine Learning for Encrypted Blockchains — Sandy Pentland, MIT

Recap of a talk with Sandy Pentland, MIT Professor, at the (Off) The Chain Summit presented by Pillar as part of Boston Blockchain Week.

Sandy Pentland, MIT Professor

Sandy Pentland is a professor at the MIT Sloan School of Management in the Information Technology group. He directs both the MIT Human Dynamics Laboratory and the MIT Media Lab Entrepreneurship program. Named one of the ‘seven most powerful data scientists in the world’, Sandy also co-leads the World Economic Forum Big Data and Personal Data initiatives. He joined us at the (Off) The Chain Summit presented by Pillar to highlight the research and projects he’s leading around machine learning for encrypted blockchains. Highlights of his talk can be found below.

Big Data is Backwards

Sandy Pentland realized early on that the infrastructure currently set up to deal with large amounts of data is vastly insufficient. When it comes to security, privacy, and data sharing, we’re clearly not living up to standard.

“How are we going to make an efficient civil system if everything is siloed?” questions Pentland. “Future systems have to be built differently.”

To combat current data security problems, Pentland has been researching applications of artificial intelligence in conjunction with blockchain. The systems he’s built revolve around three basic principles:

1. Share Answers Not Data

“Humans like to take everything and stick it one place,” explains Pentland. Whether a single google drive or hard drive, humans generally have few forms of backup data. Pentland compared this with the military circa 1400 when entire militias would stake out in a castle, only to have the entire militia destroyed when an invasion occurred. Obviously, this is no longer proper military strategy, and it shouldn’t be our data strategy either.

“If you have everything in one place, then you lose it — that’s the honeypot. It only takes one mistake to blow the whole thing. You want defense in depth,” Pentland explains.

Pentland asserts that data resources should be distributed, so you won’t be wiped out after a single attack. The way to do this is with permissioned data access and a query system, protecting the data in the event of an attack.

2. Log Everything on the Blockchain

An important part of Pentland’s approach is not only to include a query system, but also to understand what is being queried.

“Blockchain is beautiful for this!” exclaimed Pentland.

If someone queries, you should be able to tell what was asked, what the person’s permissions were, if they were approved, what algorithm was run, and what answer was given. By having everything logged, you have a complete understanding of what is happening with your data. Then, in the event of an attack, you would simply be able to reboot your data.

3. Never Decrypt Data

“If you ever decrypt data, they will steal it. You have to compute on encrypted data,” Pentland asserted. The diagram below to describes how it works:

Copyright Alex Pentland

Secure Multiparty Computation is the secret. Secure Multiparty Computation is used when a set of parties with private inputs wish to compute some joint function of their inputs (Sanadhya, Somitra).

Pentland uses the example of hospital data. Let’s say a community of hospitals wanted to figure out the accuracy of a certain procedure.

Hospital A adds their accuracy, adds a secret number, and sends it to Hospital B. Hospital B adds their accuracy, a secret number, and sends it to Hospital C. Hospital C adds their accuracy plus a secret number, and sends back to Hospital A. If any of these transactions were intercepted, it still would not tell you any information, as the data is encrypted. However, when Hospital A subtracts off their secret number, they will be able to tell the average accuracy of the three hospitals.

“Legally, that does not count as sharing data according to the law because none of it was interpretable, yet you can still get insights across encrypted data,” explains Pentland.

This is huge. “Now you can get insights across countries, across data holders, without exposing individual data and without disobeying either privacy or data localization laws.”

Machine Learning with Encrypted Insights

Pentland’s startup, Enigma, aims to add machine learning and AI to these encrypted insights. But how can you do machine learning on encrypted data?

“With human data, humans behave in very characteristic ways. These crashes and burns and adoption curves are all statistical regularities that you can find in encrypted data,” explains Pentland. Pentland calls this Social Physics.

The applications of social physics span many industries. In any instance where humans are making decisions, such as trading cryptocurrencies, taking out a loan, etc. the technology can provide meaningful insights, without having an entity share their specific data.

In one example, you can look at blockchains and can figure out who’s doing what. The methods are able to highlight which groups of people are working together without knowing who the specific people are. For bad actors, “if you can identify one person, you can identify the entire ring,” remarks Pentland.

In another example, countries can use the technology to identify carrousel crime and VAT abuse.

Other countries and companies can use the technology to restructure their infrastructure to combat against cyber attacks. “If you can see trends, you can see cyberattacks, and you can see them early,” says Pentland. “This is just the beginning. We don’t know where it will go, but it is a new way of thinking about things.”

Overall, this technology meaningfully changes the way we are able to interact with data. Because no personal data is shown, it’s possible to uncover trends and patterns in society that we wouldn’t otherwise be able to see.

“We live in a world where we actually don’t know much about what is going on around us,” Pentland explained. “We’re clueless about what’s really happening. With these sorts of systems you can begin to reveal it.”

Written by: Katie Mulligan

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Pillar

Venture capital doesn't have to be the dark side. Investing in unstoppable founders at @Algorand, @PillPack, @DesktopMetal, @PathAI & more. www.pillar.vc