Cloud Hosting

As banks, lenders, and other financial institutions increasingly turn towards artificial intelligence as a way to quickly analyze vast quantities of data, they're also attracting attention from US regulators.

Rulemakers' focus on AI is, in part, a result of concerns surrounding the safety and anonymity of customer data when used by this cutting-edge tech.

But a new tech collaboration between Nasdaq and Intel announced Tuesday might offer a solution. The partnership will see the trading venue adopt what's called homomorphic encryption, or HE, alongside Intel's latest-generation processors.

HE allows customers to perform AI and machine-learning computations on private data without ever having to decrypt the data itself, an especially critical tool in financial services where the safety of user data is increasingly important.

It's a technology that Intel also recently announced it's exploring with DARPA (The Defense Advanced Research Projects Agency, an arm of the Department of Defense) and will in Nasdaq's case provide the opportunity for next-generation computing to be tested in a business setting.

The ability to analyze large amounts of encrypted data using HE could lead to more efficient tools focused on anti-money laundering and fraud prevention efforts, areas that typically include sensitive data. Apart from finance, industries like healthcare, which tends to have restrictions around individuals' data, also stands to benefit from the use of HE.

HE could also help further increase the adoption of the public cloud, a growing trend on Wall Street. Concerns around data security in the public cloud would be alleviated by being able to keep the data encrypted.

But the problems seen in HE technology to date have always be ones of scale and speed, particularly relevant when a company like Nasdaq might be working with massive, enterprise-sized data sets.

Nir Peled, general manager of AI and private analytics at Intel, told Insider that HE is what's called "emerging" tech. It's new, complicated, and, until relatively recently, out of the scope of what most companies could manage.

"That means that nobody knows exactly where it's going to land, how it's going to be used," Peled said. "What's the ecosystem going to be around it? What are the use cases that will be first? It's literally like AI 10 years ago. A lot of promise, a lot of potential. But it's just starting."

That's where Intel's latest-generation processors come in. By using Intel's most advanced line of accelerators, Nasdaq said it will be able to dive further into homomorphic encryption, since the improved performance from these processors addresses the "performance tax" previously inherent to HE.

"[Homomorphic encryption] was really impractical in certain ways because of the performance you would get. So we decided to work with Intel on this to see exactly what kind of benefits we could get, and if that would help us make it more applicable. And I think the answer is yes," Nikolai Larbalestier, a senior vice president at Nasdaq overseeing cloud strategy and enterprise architecture, told Insider.

To that end, the partnership on HE between Nasdaq and Intel has been one of what Peled called "co-engineering" as the companies work to see where the tech will fit in to Nasdaq's data flow. At this point, HE isn't being used in any real-world business settings at Nasdaq, but the exchange has used it on proof-of-concept models internally.

Larbalestier said the two HE applications Nasdaq is currently exploring revolve around fraud detection and anti-money laundering efforts at the exchange.

"We have surveillance products that surveil the trading activity on behalf of brokers or on behalf of market operators. It's an interesting problem to be able to combine data sets between brokers to look for patterns that they may not otherwise see just by looking within one broker," Larbalestier said.

Homomorphic encryption, Larbalestier continued, would allow Nasdaq to do that all without exposing sensitive data like the "secret sauce of trading algorithms" to others.

The encrypted tech, meanwhile, also has anti-money laundering capabilities. Under existing US law, banks are able to share customer data with each other (with safe harbor provisions that shield them from liability) when they believe fraud or money laundering is occurring. But HE could expand the scope of data sharing, Larbalestier said, while also making that data more secure since it can remain encrypted.

The capabilities of homomorphic encryption, meanwhile, also go hand in hand with another form of technology that Nasdaq continues to actively explore: data storage and processing in the public cloud.

In October of last year, CEO Adena Friedman described cloud tech as "the future of the industry" while speaking at Business Insider's Global Trends Festival. And last April, the exchange launched Nasdaq Cloud Services and began streaming real-time market data to the public cloud.

"All of these data-sharing use cases that secure, multi-party computation can either make more secure or bring new capabilities to are all really enabled by the cloud and getting that scale of compute," Larbalestier said. "That probably would be incredibly difficult or expensive to do in a traditional on-prem model."

Cloud adoption and the development of homomorphic encryption are "highly correlated," according to Peled, as the tech will enable users to feel more comfortable having data sit outside their walls.

"The data is always encrypted and you can still compute on it. You can imagine what can be done on a public cloud. I don't need to trust anyone. I can do my work as if it were beyond my own firewalls," Peled said.

Originally posted here:
Nasdaq and Intel Announce Partnership Over New Encrypted Tech - Business Insider