Sponsored by: Capital One | Secure Data Collaboration: A Modern Enterprise Enclave Architecture

Summary

Capital One Software introduces a secure data enclave architecture that enables cross-functional analysis on sensitive enterprise data—including PII and intellectual property—without moving raw data, using tokenization and Databricks governance to maintain a comprehensive audit trail.
The three main barriers to sensitive data collaboration are security risk (privacy and governance challenges), compliance burden (CCPA, GDPR, and local regulations), and competitive risk (the concern that sharing data could expose proprietary information to partners or other business units).
This video covers how the enclave model builds on existing clean room infrastructure by adding tokenization and Databricks governance controls, enabling use cases like patient 360 views and clinical operations analytics where raw data must never leave its governance boundary.

Chapters

FAQs

What are the three main barriers to enterprise data collaboration?

The three barriers identified in this video are security risk (covering privacy and governance challenges), compliance burden (including CCPA, GDPR, and local regulations), and competitive risk (the concern that sharing data with partners or business units could expose proprietary information). Organizations often address parts of this through clean rooms, but gaps remain in how sensitive data enters that shared infrastructure.

How does tokenization protect sensitive data in the Capital One enclave architecture?

Tokenization replaces raw sensitive values—such as patient identifiers or financial account numbers—with non-reversible tokens before data enters the shared environment. This means no participant in the clean room sees the underlying PII, while the tokens still enable consistent joins and aggregate analytics across datasets from different organizations.

What business outcomes does the secure enclave architecture enable?

The architecture enables use cases such as patient 360 views for healthcare (tracking and optimizing patient care across provider and payer data), clinical operations analytics, and revenue acceleration through cross-organizational data partnerships—all previously blocked by security and compliance requirements that prevented raw data sharing.

How does Databricks governance support the secure enclave model?

Databricks provides the governance layer through Unity Catalog, which enforces access controls, maintains data lineage, and generates a comprehensive audit trail for all data operations inside the enclave. This ensures organizations can demonstrate regulatory compliance and that every data access event is logged and attributable to a specific identity.

Full transcript

[00:08] Hey everyone, welcome to this edition of the talk. We're going to talk about secure data collaboration. Uh my name is Vince Govious. I'm a director of product management at Capital One Software. We're an enterprise B2B arm of uh Capital One
[00:21] where we focus on selling uh data security to enterprises. And uh I want to touch on um you know, data collaboration. But before we talk about collaboration,
[00:34] I want to talk about, you know, the value of silo data. You You go to any organization today, uh you know, across all sectors. You're talking about um you know, doesn't matter which
[00:46] industry, which vertical, highly regulated, uh government, financial institutions. They all have sensitive uh data that requires for them to drive
[00:58] business outcomes. Now, what happens when you join that data with other data sets to provide powerful analytics? That is an unlock for us. And there's a reason why there's silo
[01:11] data because they have to adhere to regulations, to data sovereignty. There's also uh you know, strategic directions that they have to do to make make that data visible only within the organizations. However, there are also barriers to
[01:25] collaboration. How do you collaborate with these silo data? Organizations typically talk about, you know, three aspects. There's a security risk. You're talking about, you know,
[01:37] privacy. Uh there's security aspects, there's governance challenges, there's also organization nuances around what you can or cannot share. There's also a compliance burden, right? You want to
[01:49] talk about CCPA, GDPR mandates. There's also local uh rules and regulations that you need to adhere to. That's the compliance aspect of that you have to deal with. There's also the competitive risk where,
[02:02] you know, you fear leaking proprietary information when you're trying to share data, right? All of these are barriers to entry for collaboration. Now, in terms of the, you know, what
[02:14] organizations have done, at least what we are seeing in the industry today, is they've introduced the concept of clean rooms. Now, in clean rooms, what happens is they have built the underlying infrastructure. They've built controls.
[02:28] They've built, uh, you know, collaborative tools. They have infrastructure to be able to facilitate a mandate. However, there's still that barrier to entry for, you know, how do I put my sensitive data in
[02:40] this infrastructure? There is still, uh, a gap that we have to deal with. So, what is the opportunity here, right? So, when you're talking about opportunity, we're talking about how do I securely
[02:53] get data in to the environment, right? How do I make sure the data I'm using to collaborate with is secure enough so that I do not violate my local,
[03:07] uh, you know, regulations and compliance mandates that are required for this data. And that is the opportunity here. That is the opportunity. How do you, you know, collaborate without exposing the raw elements? Now, in Capital One, we
[03:21] focused on three principles. Now, oops.
[03:34] Sorry. All right. So, in Capital One, we focused on three principles. The first one is privacy-preserving data collaboration. What this means is I want to be able to collaborate with your data without exposing the underlying
[03:46] sensitive elements of that data. That is a That's the number one principle. The second principle is control. You want to be able to exercise control over that data. It doesn't matter whether you are
[03:58] you have your own governance policies and that data you're trying to share with other third-party organizations, but you want to maintain control of that data. That's the second principle. The third principle, and which is the most
[04:10] important, is simplicity. Now, we all know we tend to over-engineer and provide excessive guardrails around data. However, that's also a hindrance and barrier to entry for accessing that
[04:22] data, especially for analysts and data scientist community. So, that is the third aspect that we dealt with, which is simplicity. Now, these were the fundamental reasons why we uh built such a uh modern collaboration
[04:35] tool. So, now what did we uh you know, realize through these uh outcomes, right? So, we're trying to build better customer experiences. We're trying to streamline efficiency. We're trying to drive strategic business
[04:48] growth. So, what does it mean? What does it mean to us, right? So, let me walk you through, you know, building better customer experiences. Let's say, for example, you want to build a holistic view of uh
[05:00] a patient, right? You want to build a 360° view of your patient. How do you do that? Because you do not have all the data sets. If you can collaborate and you get, you know, their entire data sets linked without having to expose
[05:13] their sensitive elements, you can build that 360° view. But, what is that going to drive for you? That is going to drive personal care. It's going to drive improved outcomes for that patient. It's going to drive um you know, improved
[05:26] health prognosis, right? Improved care, improved research. You cannot do that without collaboration of data, because no one organizations holds the keys to these data sets unless you collaborate.
[05:38] So, that's the first outcome, right? The second one is efficiency. Now, you know, in this example, we talk about an analyst and a scientist. An analyst wants to, you know, run, you know, analytical models. They want to
[05:51] track and optimize patient care. Again, a healthcare use case. However, it requires a lot of manual data prep, right? You We want to eliminate that. Uh the same thing for a scientist. They
[06:03] want to, you know, do a lot of operational clinical, you know, mandated data outcomes. How do they do that? You know, without uh having to go through mandated access controls. How do you get
[06:15] faster time to insights? This is the way to do it. So, what is the outcome that we're driving here? You're trying to reduce your data prep overhead from months, right? And also, you want to accelerate discovery and time to insight
[06:27] for that data. The last thing which all organizations care about is accelerated revenue opportunities. Now, in this, you want to drive revenue growth for your company. How do you do that? You accelerate that
[06:39] by collaboration, right? Everybody talks about innovation. How do you speed up innovation? Again, collaboration. You are um you know, speeding up the uh need for innovation, and we're seeing every
[06:51] industry does is doing that today. Partnerships. We all know that we cannot do it all alone. We have to partner. We have to collaborate. And that is what this uh helps us drive towards. Now, in terms of
[07:04] uh strategic advantage, we are talking about secure data collaboration is no longer barrier here, right? What What do we mean by um you know, reducing that barrier to
[07:16] entry? You no longer have to worry about exposing your data risk if you adhere to these three principles, right? You can make it as a competitive advantage. Now, to that aspect, I want to introduce
[07:28] Databricks Connect, right? This is the newest uh product in our portfolio at Capital One Software that we have introduced jointly with Databricks. Uh it was announced It was the first mention in our keynote on Tuesday.
[07:42] And uh this is jointly uh published on the app marketplace that was launched with Databricks. It It is fundamentally built to solve the data collaboration problem. Now, in
[07:54] this example, you see two collaborators. There's a collaborator A and a collaborator B. Both have distinct data sets. Both have sensitive elements that they do not want to expose. However, you can tokenize that data
[08:07] and bring those tokenized data sets into a clean room environment. So, using a shared clean room infrastructure, Databricks provides you that framework to be able to lock down the role-based access controls, your infrastructure
[08:20] security, all aspects of it. So, Databricks takes care of it. However, we still protect the data. So, that data is still tokenized. Now, the beauty of this implementation is you can link these
[08:32] tokens. So, you do not have to re-identify these tokens anymore. So, you can link these data sets and you collaborate and you can get matches on these tokens. So, that is the value proposition that this tool provides.
[08:45] Now, in addition to that, you also can talk about, "Hey, what about the boundary? Am I exi- exiting my org boundary?" Absolutely not. This stays within your Databricks account. It is
[08:58] within your organizational boundary. When it comes to data prep, you're using your own native Databricks compute. When it comes to linking, you're using the Databricks compute within the clean room infrastructure. So, all within your account. So, you do not cross your
[09:11] organizational boundary for data collaboration. Of course, you want to invite people into your clean room, that's a uh you know, data sharing agreement that you have with a third party, but that is regardless of how,
[09:24] you know, the infrastructure is set up. But, this is the you know, the premise of adding this capability. This is a new ascent rate. We've had tremendous feedback and uh
[09:36] terms of enabling and unlocking value for your organizations, this is the tool that uh we recommend you take a look at. You know, you can go to the App Store and uh
[09:49] try it out and let us know what you think. And uh that's it for my presentation. Uh thank you for joining. I'm happy to take any questions. I'll be here with my colleagues if you have any
[10:02] questions, but uh thank you and enjoy the rest of the conference.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.