Explore the nature of vulnerabilities in this episode of ThreatWise TV.
It’s shaping up to be another big year for vulnerability disclosure. Already the number of Common Vulnerabilities and Exposures (CVEs) disclosed has crossed 18,000 and it’s on track to make this another record-breaking year.
With new CVEs being disclosed daily, it has become increasingly difficult for security teams to stay abreast of the latest risks, let alone quickly determine which ones apply to their network environment. From those, prioritizing which CVEs to patch first adds an additional wrinkle to the process.
If this wasn’t challenging enough, a curve ball that’s often lobbed at security teams are the “breaking news” vulnerabilities— vulnerabilities picked up by the security media, often with much fanfare. The stories surrounding these high-profile vulnerabilities generally carry an implied threat that the CVE in question will throw the doors wide open to attackers if not addressed immediately. What security team hasn’t had someone from the C-suite share an article they’ve read, asking “are we protected from this?”
On the surface, CVEs that appear severe enough to garner media attention do seem like a good place to start when addressing vulnerabilities in your environment. But vulnerabilities are complicated, and what a security researcher manages to do within a controlled environment doesn’t always translate into real-world attacks. In fact, most disclosed vulnerabilities never see active exploitation. And of those that do, not every vulnerability ends up becoming a tool in an attacker’s arsenal. Bad actors generally follow the path of least resistance when they compromise a network, relying on tested exploits long before trying something new and unproven.
This begs the question: how much overlap is there between the most talked about vulnerabilities and those that are widely used in attacks? Moreover, if media attention isn’t a reliable indicator, what else might predict if a vulnerability will be used in an attack?
How to compare exploitation and media attention
To answer these questions, we used intelligence tools available from Cisco’s Kenna Security risk-based vulnerability management (RBVM) software. In particular, Kenna.VI+ consolidates a variety of vulnerability intelligence, where a CVE ID lookup can pull back a wealth of information. In addition to this, Kenna.VI+ includes an API that brings in an additional layer of external threat intelligence, enabling further analysis.
We started with a direct comparison of Successful Exploitations and Chatter Count from within Kenna.VI+. The former is a full count of confirmed exploits within the dataset, while the latter is a count of mentions in the news, social media, various forums, and the dark web.
The 30,000-foot view
Our first pass at the data included a comparison of the top 50 CVEs in both Successful Exploitations and Chatter Count. However, there were only two CVEs that overlapped. The data showed that many of the top exploited CVEs were old and predated the data in Chatter Count. We quickly decided that this wasn’t a fair comparison.
To get a better look at more relevant CVEs, we limited the dataset to a range of 10 years. Unfortunately, this did not do much to improve things—only three CVEs showed up in both lists.
The wheat from the chaff
A more effective approach was to look at CVEs that we know are actively being exploited. The Cybersecurity and Infrastructure Security Agency (CISA) happens to maintain such a list. The Known Exploited Vulnerabilities (KEV) catalog is considered an authoritative compilation of vulnerabilities identified as being actively exploited in the wild.
Running the KEV catalog though Kenna.VI+ resulted in six CVEs that appeared in the top 50 for both lists, with a single overlap in the top 10. This leads us to conclude that the vulnerabilities with the most discussion are not the same as those being actively exploited in the majority of cases.
Top 10 successfully exploited CVEs
|1||CVE-2017-9841||PHPUnit vulnerability (used to target popular CMSes)|
|3||CVE-2019-0703||Windows SMB information disclosure vulnerability|
|5||CVE-2017-9805||REST plugin in Apache Struts vulnerability|
|6||CVE-2017-11882||Microsoft Office memory corruption vulnerability|
|7||CVE-2017-5638||Apache Struts vulnerability (used in Equifax breach)|
|8||CVE-2012-1823||10-year-old PHP vulnerability|
|10||CVE-2018-11776||Apache Struts RCE vulnerability|
Top 10 most talked about CVEs
|1||CVE-2021-26855||Microsoft Exchange vulnerability (used in Hafnium attacks)|
|2||CVE-2021-40444||Microsoft MSHTML RCE vulnerability|
|3||CVE-2021-26084||Confluence Server and Data Center vulnerability|
|4||CVE-2021-27065||Microsoft Exchange vulnerability (used in Hafnium attacks)|
|5||CVE-2021-34473||Microsoft Exchange vulnerability (used in Hafnium attacks)|
|6||CVE-2021-26858||Microsoft Exchange vulnerability (used in Hafnium attacks)|
|8||CVE-2021-34527||One of the PrintNightmare vulnerabilities|
|9||CVE-2021-41773||Apache HTTP Server vulnerability|
|10||CVE-2021-31207||One of the ProxyShell vulnerabilities|
Name recognition on both sides
Despite the lack of overlap, there are many well-known vulnerabilities at the top of both lists. Heartbleed and EternalBlue appear on the top 10 exploited list, while Hafnium, PrintNightmare, and ProxyShell make the top 10 most talked about CVEs.
The Log4j vulnerability is the only CVE that appears in both lists. This isn’t surprising considering the ubiquity of Log4j in modern software. It’s the second-most exploited vulnerability—far outpacing the CVEs directly below it. This, coupled with its appearance in the chatter list, puts it in a class of its own. In a brief period, it’s managed to outpace older CVEs that are arguably just as well known.
The CVE that recorded the most successful exploitations is a five-year-old vulnerability in PHPUnit. This is a popular unit-testing framework that’s used by many CMSes, such as Drupal, WordPress, MediaWiki, and Moodle.
Since many websites are built with these tools, this exploit can be a handy vector for gaining initial access to unpatched webservers. This also lines up with research we conducted last year, where this vulnerability was one of the most common Snort detections seen by Cisco Secure Firewall.
All four of the Microsoft Exchange Server vulnerabilities used in the Hafnium attacks appear in the most talked about list of CVEs. However, even when you add all four of these CVEs together, they still don’t come anywhere close to the counts seen in the top exploited CVEs.
If media attention is not a good predictor of use for exploitation, then what are the alternatives?
The Common Vulnerability Scoring System (CVSS) is a well-known framework for gauging the severity of vulnerabilities. We looked for CVEs from the KEV catalog that were ranked as “critical”—9.0 and above in the CVSSv3 specification. Examining the entire KEV catalog, 28% of the CVEs have a score of 9.0 or higher. Of the top 50 successfully exploited, 38% had such scores.
This is an improvement, but the CVSSv3 specification was released in 2015. Many CVEs in the KEV catalog predate this—19% of the entire catalog and 28% of the top 50—and have no score.
Using the previous CVSS specification does fill this gap—36% overall and 52% of the top 50 score 9.0 or higher. However, the older CVSS specification comes with its share of issues as well.
Another indicator worth exploring is remote control execution (RCE). A vulnerability with RCE grants an attacker the ability to access and control a vulnerable system from anywhere. It turns out that 45% of the CVEs in our dataset allow for RCE, and 66% of the top 50, making it the most worthwhile indicator analyzed.
Honing the approach
Let’s summarize how we’ve honed our approach to determine if media attention and exploitation line up:
|Data set||Exploitation and Chatter lists||Number of CVEs|
|All CVEs||Appears in both top 50||2|
|Appears in both top 50 (last 10 years)||3|
|KEV Catalog||Appears in both top 50||6|
|Appears in both top 10||1|
And here’s a summary of our look at other indicators:
|KEV Catalog||Top 50 exploited|
|Allows for RCE||45%||66%|
All of this analysis provides a clear answer to our original question—the most regularly exploited CVEs aren’t the most talked about. Additional work highlights that monitoring variables like RCE can help with prioritization.
For illustrative purposes we’ve only looked at a few indicators that could be used to prioritize CVEs. While some did better than others, we don’t recommend relying on a single variable in making decisions about vulnerability management. Creating an approach that folds in multiple indicators is a far better strategy when it comes to real-world application of this data. And while our findings here speak to the larger picture, every network is different.
Regardless of which list they appear on, be it Successful Exploitations or Chatter Count, it’s important to point out that all these vulnerabilities are serious. Just because Hafnium has more talk than Heartbleed doesn’t make it any less dangerous if you have assets that are vulnerable to it. The fact is that while CVEs with more talk didn’t make the top of the exploitation list, they still managed to rack up tens of thousands of successful exploitations.
It’s important to know how to prioritize security updates, fixing those that expose you to the most risk as soon as possible. From our perspective, here are some basic elements in the Cisco Secure portfolio that can help.
Kenna Security, a pioneer in risk-based vulnerability management, relies on threat intel and prioritization to keep security and IT teams focused on risks. Using data science, Kenna processes and analyzes 18+ threat and exploit intelligence feeds, and 12.7+ billion managed vulnerabilities to give you an accurate view of your company’s risk. With our risk scoring and remediation intelligence, you get the info you need to make truly data-driven remediation decisions.
To responsibly protect a network, it’s important to monitor all assets that connect to it and ensure they’re kept up to date. Duo Device Trust can check the patch level of devices for you before they’re granted access to connect to corporate applications or sensitive data. You can even block access and enable self-remediation for devices that are found to be non-compliant.
How about remote workers? By leveraging the Network Visibility Module in Cisco Secure Client as a telemetry source, Cisco Secure Cloud Analytics can capture endpoint-specific user and device context to supply visibility into remote worker endpoint status. This can bolster an organization’s security posture by providing visibility on remote employees that are running software versions with vulnerabilities that need patching.
Lastly, for some “lateral thinking” about vulnerability management, take a look at this short video of one of our Advisory CISOs, Wolfgang Goerlich. Especially if you’re a fan of the music of the 1920s…
We’d love to hear what you think. Ask a Question, Comment Below, and Stay Connected with Cisco Secure on social!
Cisco Secure Social Channels