How Snowflake’s cloud data lake can be used for security
For a startup in hyper-growth mode like Figma, scaling up its cybersecurity defenses as fast as the rest of the company has been a top concern.
To pull off that feat, the company has come to rely on Snowflake, a company well known for its cloud data lake and data warehouse technology, but much less so for what it brings to the table for cybersecurity. As it turns out, though, "the same reason everyone else is using Snowflake and finds its capabilities so powerful also applies to security," said Devdatta Akhawe, head of security at Figma.
Snowflake’s technology is primarily used for cloud-based data analytics and data science, but it is now looking to prove it has a lot to offer when it comes to cybersecurity, which is increasingly being recognized as a data problem at its core.
Figma — which offers browser-based, collaboration-oriented design software, and is on tap to be acquired by Adobe for $20 billion — believes Snowflake has already arrived in the cybersecurity market given its unique ability to combine security feeds with data from the rest of the business. Having a single data lake for the whole company has allowed for analysis of cybersecurity data in a broader context, enabling a better understanding of security risks, according to Akhawe.
"Being able to correlate across a large number of disparate data sources is what makes a strong security program," he said.
Snowflake executives told Protocol that while cybersecurity is just one of the cloud data opportunities the company is pursuing right now, it's clearly among the biggest. The company's emphasis in the space comes amid intensifying cyberthreats facing enterprises and a growing priority placed on cybersecurity in the C-suite and boardroom.
The relevance of Snowflake for security teams is "the best-kept secret in cybersecurity," said Omer Singer, the company’s head of cybersecurity strategy.
Many of the initial customers using Snowflake for cybersecurity — which include Dropbox, DoorDash, TripActions, and CSAA Insurance Group — have been "using Snowflake for a long time, but the cybersecurity team wasn't," Singer said. "What's changed is, now the cybersecurity team is using it as part of their overall strategy."
Correlating threats
Founded in 2012 by two veterans of Oracle — Benoit Dageville (now president of product) and Thierry Cruanes (now CTO) — Snowflake has homed in on cybersecurity in the three years since Frank Slootman joined as CEO, said Christian Kleinerman, senior vice president of product at Snowflake.
Early on at Snowflake, Slootman, who was formerly the CEO of ServiceNow, had a decisive meeting with a customer, according to Kleinerman. "The customer was telling him, 'We are doing cybersecurity with Snowflake — why are you guys not pitching it as such?'"
In June, Snowflake announced its new cybersecurity category, aiming to offer an easier way for customers to combine their security data with other business and contextual data.
Doing so can enable better-informed threat detections and breach investigations, according to Snowflake executives. For instance, correlating human resources data with email-forwarding events to outside parties could help detect if an employee is trying to leak sensitive information.
The more signals you have, the more patterns you can find.
Meanwhile, combining data feeds from code repository GitHub and identity platform Okta could provide a view into who's logging into privileged accounts, what they're doing, and whether permissions violations are taking place, according to Snowflake executives. Typically, however, suspicious developer behavior would be missed, because GitHub data isn't usually leveraged by security operations tools or staff members.
"The more signals you have, the more patterns you can find," Kleinerman said.
That's not something you can easily do with traditional data warehousing technology, however. For one thing, cybersecurity is different from other parts of a business in that it generates way more data — a nonstop stream of logs and events. For customers, storing security data for any period of time has often been costly and required tough choices about what to keep.
Snowflake's separation of pricing between storage and compute, however, "works very well for security," said Uri May, co-founder and CEO of cybersecurity vendor Hunters. In security, you want to store a lot of data for potential review later on, he said, but you probably won't need to have access to all of your data all of the time.
Using Snowflake, though, an organization only pays for compute time on its security data when an incident has actually occurred and the stored data needs to be queried, May said. The rest of the time, you're just paying the "relatively low" price for the storage itself.
By contrast, customers that try to store security data using a system that predated the cloud — which doesn't separate storage from compute, and doesn't leverage cloud-native storage architecture — will be forced to get choosy about what data they collect and how long they keep it, Singer said.
That's not great for security, said Figma's Akhawe. As occurred with the widely felt SolarWinds attack, many high-impact breaches are only discovered nine months to a year after they first began, he said.
Deleting security data after a few months "is just illogical. You're flying blind when the actual breach is disclosed," Akhawe said. Snowflake, on the other hand, "gives us the ability to scale to gigantic amounts of [security] data."
Bringing apps to the data
Snowflake executives said that the company is encouraging third-party software vendors to provide the security features around its data platform. "Instead of taking the data to applications, let's bring the applications to the data," Kleinerman said.
Vendor partners include Hunters, which provides security analytics and correlation for data in Snowflake; Immuta, which offers access control and privacy management; and Lacework, which focuses on enabling threat detection, investigations and measurement of security, and compliance posture.
Providing customers with a way to have all of their data in one place "gives them a holistic view of what's happening in their business, with security becoming more and more important to every business," said David Hatfield, co-CEO at Lacework, which has received an investment from Snowflake and was incubated at the same private equity firm, Sutter Hill Ventures.
In the case of Dropbox, the company has transitioned from using a traditional platform for security information and event management to using a cloud-native SIEM from Panther Labs. "They have completely decommissioned their traditional SIEM," Singer said.
The use of Snowflake for cybersecurity is just starting to move from early adopters to more mainstream usage, executives said. Cybersecurity is one of the eight categories currently being promoted for the platform, but it is the first to target a specific audience within an enterprise.
Two other new categories, planned to be announced in 2023, will similarly target a more specific audience, though details aren't being disclosed for now, Singer said. Previously, Snowflake categories have been more general in nature, targeting uses such as data science and data engineering.
‘Massive opportunity’
In terms of the cybersecurity push, "I think our opportunity to do something meaningful in this space is massive," Kleinerman said.
Snowflake's emerging focus on security comes up in his discussions with customers on a weekly basis, he said. "I can tell you, it's a conversation changer."
Without a doubt, it's easier for younger companies that can start fresh with their data architecture, like Figma, to adopt Snowflake for their cybersecurity needs, Akhawe said.
"I think for a lot of other companies that have legacy architecture, it is harder to migrate," he said.
But while it will take time for companies to make the shift, Akhawe believes that 10 years from now, most will have transitioned to this type of architecture for their security data.
"We know attacks are getting more sophisticated and more complicated. And very often, they take longer than three or four months to detect," he said. As a result, using a large-scale, cloud-based data lake "where you don't have to worry about deleting data — I do think this will become the default."
Gloss