Cybersecurity hype keeps building around XDR. So does confusion.
Proponents say that extended detection and response services have huge potential for improving security for customers – if only top industry players could agree on what XDR actually is.
In mid-2018, Nir Zuk, the founder and CTO of Palo Alto Networks, took the stage at a company event and introduced the world to a new type of cybersecurity product. In the four years since, his concept, which he dubbed "XDR," has swept through the industry. It's now a focal point for virtually every major vendor in the security industry.
In a recent interview, Zuk did not sound happy about the whole thing. Not at all.
XDR, which stands for "extended detection and response," revolves around the premise that security is most effective when all the data from across a customer’s IT environment can be correlated and analyzed together as a whole. It aims to accomplish this feat by bringing together all of a customer's systems and cybersecurity tools into a unified, integrated platform.
Certainly, the cybersecurity industry is notorious for its buzzwords and acronyms. But XDR is not your average security acronym: If you believe the leaders of many top players in the industry, XDR could be the architecture of the future for cybersecurity.
XDR is "the way to actually prevent damage from breaches, and the way to scale and deeply automate security with a scarce talent pool," said Wendy Thomas, president and CEO of Secureworks, which heavily focuses on XDR.
According to proponents, embracing an XDR-based approach can address many of the pressing issues that security teams face: the overload in alerts, difficulty in prioritizing threats, tool sprawl. As IT gets more complex, "it's becoming harder and harder for humans to operate cybersecurity," Zuk said.
However, he is far from thrilled with how others have adapted his idea. There are too many varying uses and misuses out there right now for “XDR” as a term — and in many cases, it's just a new label slapped on old products, Zuk argued.
"I think XDR, today, is just a term that different vendors use differently," he said, acknowledging that XDR has joined a long line of enterprise tech terms that have devolved into nebulous buzzwords.
Still, the high level of attention around XDR makes the confusion in the market a bigger issue than it might normally be. XDR is expected to see surging adoption in the coming years, with Gartner forecasting that 40% of organizations will be deploying the technology by 2027, up from 5% as of last fall.
"XDR is definitely something that we recommend organizations look into," said Patrick Hevesi, a vice president and analyst at Gartner, thanks to its ability to integrate more data feeds into detection and response efforts.
Thinking like a hacker
XDR aims to detect security issues across entire IT environments because that's how attackers operate: Hackers get inside one system operated by an organization, then move around to others during the course of an attack.
As the thinking goes, if you just look at the endpoint — or network, application or cloud infrastructure — you're only going to see a slice of what an attacker is doing. If you can view everything together, as XDR seeks to, then you have a better shot at stopping attacks such as ransomware at an early stage.
In other words, XDR is the security industry's answer to many of the questions that customers are asking as they grapple with an increasingly complex set of environments in 2022.
Most XDR vendors agree on all these reasons for why the approach is so promising. But from there, the question of how to define XDR gets more contentious.
"I firmly believe it's one of the most misused or abused terms in the industry," said Michael Sentonas, CTO at CrowdStrike, which made its name on endpoint detection and response (EDR) and announced its expansion into XDR last fall.
It's notable that so many of the biggest players in the industry are moving aggressively to offer some version of an XDR platform. In addition to CrowdStrike and Secureworks, Microsoft, SentinelOne, Mandiant, Trellix, Sophos, Cisco and VMware are among those who've joined Palo Alto Networks on the list of XDR platform vendors.
As of this writing, Protocol has identified 34 security vendors that are marketing XDR products, and there are likely many more. (The figure also excludes providers of managed XDR services.)
But at this stage, there is little agreement among industry players about what constitutes a "true" XDR — leaving it up to customers to figure out what's what.
"It's hard to talk about the term ‘XDR,’ because every organization defines it the way that they want. And industry analysts have not yet solidified, as a group, what the definition of XDR is," said Mandiant CTO Marshall Heilman.
Extending to new areas
When Zuk first revealed his notion of XDR four years ago, he chose the terminology to make a specific point: EDR, or endpoint detection and response, is not sufficient because attackers don't just target the endpoint. The same problem applies to detection tools just focused on the network, cloud or applications. In Zuk's original definition of XDR, the "X" stood for "anything" — as in, any type of system that a threat actor might leverage in an attack.
However, according to the consensus today, the “X” stands for "extended." As in, detection and response that extends past any one environment.
The industry's conception of XDR has also evolved in ways that are more consequential. For one thing, many vendors now offer XDR-branded products that also leverage data from third-party tools.
Platforms that use data from multiple vendors' tools are now commonly referred to as "open" or "hybrid" XDR. Offerings that use data from a single vendor's tools — such as Palo Alto Networks, Microsoft or Cisco — have come to be known as "native" XDR.
Both native and open XDR approaches can have their advantages, though much depends on how the vendor sets things up, said Forrester Analyst Allie Mellen. Open XDR is touted as offering the flexibility to leverage existing security tools, bringing together data feeds from the products that customers have already invested in.
But that apparent advantage of open XDR could actually be a downside if the third-party tools are not integrated effectively, Mellen said.
"I question whether or not the detection quality is going to remain high, if they're just developing integrations willy-nilly and trying to support as many as possible," she said. The purpose of XDR is to better tailor and curate the experience for the security team, so managing integrations well is critical. "It has to be done with intentionality," Mellen said.
Ultimately, she is hesitant to say that open XDR is generally superior to native XDR. Some native XDR vendors have gained reputations for providing high-value detections, "because they know and understand everything in the environment," Mellen said. "They know all of the telemetry that's coming in. And they choose what telemetry is coming in. So I think it's a bit of a trade-off."
A related issue with open XDR is that, essentially, no single vendor is accountable for the security outcome from the use of the platform, said Frank Dickson, group vice president for Security and Trust at IDC.
If a customer chooses to secure their environment with an XDR platform that ties together tools from disassociated vendors via APIs, then the customer is accountable, Dickson said.
"That's one of the shortcomings of open XDR," he said. "By open XDR, what that fundamentally says is, the customer owns the outcome. The vendor doesn't own the outcome."
'Not being honest'
Zuk argues that there's an even bigger problem with open XDR. Such platforms give the impression that they can leverage data that they don't actually have, he says.
For instance, Palo Alto Networks is among the largest network security vendors, but "none of these [open XDR] vendors is using our data," Zuk said.
Ultimately, open XDR vendors "are not being honest when they say that they have the third-party data," he said. All of which means that the results for attack detection and response are inevitably going to be "sub-optimal," according to Zuk.
An executive at another prominent native XDR vendor, Microsoft, made a similar point. Rob Lefferts, corporate vice president for Microsoft 365 Security, said that to be effective with XDR, "you have to actually deeply know the tool that you are investigating — you can't just dump in a bunch of data."
To Lefferts, the concept of open XDR seems to be no different than that of security information and event management (SIEM). And indeed, analysts have noted that a number of SIEM vendors have simply rebranded their products as XDR.
"I look at open XDR, and I'm like, 'Oh, you mean a SIEM? Is that what an open XDR is?'" Lefferts said.
Not surprisingly, executives at major providers of open XDR platforms would disagree.
While some open XDR platforms do have their origins as a SIEM, that's not universally true, said Secureworks Chief Product Officer Steve Fulton. His company touts its open XDR platform as being "purpose-built" for running detection and response across multiple environments.
Open XDR recognizes that most customers do not have tools from just one vendor in their environment and will prefer to leverage their existing investments, Fulton said. Most customers do not want to have to "rip and replace" their security tool set just to use XDR, according to Fulton.
"If you're a native XDR vendor and you're saying, 'You have to have our stack in order to get value out of XDR,' you're going to be pretty narrow in your scope. You're going to miss some things with that approach," Fulton said. "My view is, that approach is going to die away over time."
With an open approach to XDR, on the other hand, "we firmly believe it drives the best security outcomes for our customers," he said. "It's going to give you the widest possible aperture."
At Trellix, the company formed through the merger of McAfee Enterprise and FireEye, CEO Bryan Palma pointed to XDR as the vendor's biggest opportunity looking ahead. The company's core strategy following its rebranding announcement in January has been seeking to become the leading player in XDR, in fact, by integrating both native tools and an open XDR approach.
XDR is "not next-gen SIEM. It's not next-gen endpoint [security]. It's broader. It's a platform,” Palma said. "It's bringing together your capabilities to create a next-level architecture — which is very different than, 'This is the next SIEM.’”
That being said, there are certain elements that any XDR provider should be expected to offer, namely, endpoint detection and security operations capabilities, according to Palma.
"I just don't know how you're a viable player [in XDR] if you don't have an endpoint," he said. "I think to be a true XDR, you've got to have endpoint capabilities."
In terms of security operations, XDR should be able to automatically correlate security issues detected across different environments and present the findings to security analysts for further investigation, Palma said.
CrowdStrike's Sentonas goes a step further in his criteria for what constitutes true XDR: Not only should XDR be able to cut across all of a customer's environments, but there should be essentially no difference between the data coming in from different vendors' tools, allowing for detections that work effectively regardless of the data source.
"The problem we believe we should be solving with XDR is not to just bring in third-party data, but to actually do something meaningful with it, and that is to focus on automated detections," Sentonas said.
Doing this entails a concerted effort around ensuring that the security data "all works the same. It all looks the same. The language between all the vendors, if you will, is exactly the same," he said. The benefit is that machine learning models "should work the same way on another vendor's data as it does on ours."
Many XDR vendors, however, are not treating the data in this way, so they can't extend all of their native detection and response capabilities to third-party tools, Sentonas said.
A number of XDR vendors fall short when it comes to the "response" portion of extended detection and response, according to Nicholas Warner, president of Security at SentinelOne.
"That is the difference between XDR and SIEM," Warner said. "Anyone can generate an alert. Not just anyone can actually orchestrate a response — and then make that an automatic response and an effective response."
And that is where endpoint detection and response vendors have a natural advantage over "pure-play" XDR vendors, he said.
"In which way could a pure-play XDR vendor do execution control on a system as a response? And the answer would be, they wouldn't be able to," Warner said. "And that's pretty big. Because that is the 'R' in XDR."
Marketing about XDR was ubiquitous at the RSA Conference in San Francisco earlier this month. The only serious rival for the biggest buzzword at the conference was "zero trust" — which, like XDR, lacks an agreed-upon definition by the security industry.
But compared to zero trust, which is widely understood to be more of an architectural concept, XDR is actually a product in some cases. Increasingly, it's also being offered as a managed service, given the shortage of available security professionals to operate an XDR platform.
Andrew Maloney, co-founder and COO at cybersecurity firm Query.AI, which does not offer an XDR platform, said he thinks the idea of taking down silos between data and tying all systems together is the right goal for cybersecurity as a whole.
But whether you're talking about a "native" or "open" approach to XDR, "the buzz is outpacing the market," Maloney said.
"Now every big player claims an XDR capability,” he said, “whether they have it or not."