If you are reading this blog post via a 3rd party source it is very likely that many parts of it will not render correctly (usually, the interactive graphs). Please view the post on dogesec.com for the full interactive viewing experience.
tl;dr
The CVE List was launched in September 1999, listing 321 CVE records. 25 years later there are 265,767 CVE records.
Key Findings
- 2024 already holds the records for the most vulnerabilities published in a single year
- RedHat (11,565), Microsoft (11,030), and Oracle (8,042) are the CNAs that have submitted the most vulnerabilities for their products.
- The mean CVSS base score is 6.5
- Most CVEs are published on a Wednesday (54,862, 21.3%)
- Linux operating systems have the most published vulnerabilities
Why I decided to write this post
It is a homage to the CVE program.
We all owe a lot to the people behind it.
Follow along
For this post I am going to use the STIX Objects created by cve2stix over this period.
If you would like to follow along, you can download the same dataset I will be using with the help of stix2arango.
Once you have installed stix2arango, all you need to run is;
python3 utilities/arango_cti_processor/insert_archive_cve.py \
--database cti_knowledge_base_store
This command will import all CVEs published by the NVD into the graph database (including those since this post was written), ready for querying in the ArangoDB web interface (usually http://127.0.0.1:8529/).
Analysis
Count of CVE records
This query counts all Vulnerability objects
RETURN LENGTH(
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
RETURN doc
)
[
265767
]
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
COLLECT year = SUBSTRING(doc.created, 0, 4) WITH COUNT INTO count
SORT year
RETURN { year: year, count: count }
With 3 months left of the year, this year already holds the records for the most vulnerabilities published (31,139).
Most active CNAs
This search groups all Vulnerabilities by the CNA that submitted them
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
FOR ref IN doc.external_references
FILTER ref.source_name == "sourceIdentifier"
COLLECT description = ref.description WITH COUNT INTO count
SORT count DESC, description
RETURN { description: description, count: count }
MITRE, the Root CNA responsible for managing the CVE program, top the list, which is to be expected. As is the fact the rest of the top 10 is made up of major software vendors, RedHat, Microsoft, Oracle, etc.
There has been a lot of complaints in the community about poorly written submissions, or put more bluntly, submissions that should never had been made in the first place.
This search is similar to the previous one, however, this time counts the number of CVEs rejected by the NVD by CNA;
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND (
LENGTH(
FOR ref IN doc.external_references
FILTER ref.source_name == "vulnStatus"
AND ref.description == "Rejected"
RETURN ref
) > 0
)
FOR ref IN doc.external_references
FILTER ref.source_name == "sourceIdentifier"
COLLECT description = ref.description WITH COUNT INTO count
SORT count DESC, description
RETURN { description: description, count: count }
I’m looking at you, Intel!
CVEs by CVSS base score
The CVSS base score reflects the inherent severity of a vulnerability before any mitigation efforts are considered, providing a consistent way to compare vulnerabilities.
Here is a search that shows the distribution of results by CVSS base score;
LET filtered_docs = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND NOT (
LENGTH(
FOR ref IN doc.external_references
FILTER ref.source_name == "vulnStatus"
AND ref.description == "Rejected"
RETURN ref
) > 0
)
RETURN doc
)
LET base_score_groups = (
FOR doc IN filtered_docs
// Iterate over each cvss version
FOR cvss IN VALUES(doc.x_cvss)
LET base_score = cvss.base_score
FILTER base_score != null
// Classify the base_score into different ranges using ternary operator
LET range =
base_score >= 0 AND base_score < 1 ? "0-1" :
base_score >= 1 AND base_score < 2 ? "1-2" :
base_score >= 2 AND base_score < 3 ? "2-3" :
base_score >= 3 AND base_score < 4 ? "3-4" :
base_score >= 4 AND base_score < 5 ? "4-5" :
base_score >= 5 AND base_score < 6 ? "5-6" :
base_score >= 6 AND base_score < 7 ? "6-7" :
base_score >= 7 AND base_score < 8 ? "7-8" :
base_score >= 8 AND base_score < 9 ? "8-9" :
base_score >= 9 AND base_score <= 10 ? "9-10" : "unknown"
COLLECT score_range = range INTO group
RETURN {
score_range: score_range,
count: LENGTH(group)
}
)
// Sort by the score range
FOR entry IN base_score_groups
SORT entry.score_range
RETURN entry
The mean and median scores can be calculated using this search;
LET filtered_docs = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND NOT (
LENGTH(
FOR ref IN doc.external_references
FILTER ref.source_name == "vulnStatus"
AND ref.description == "Rejected"
RETURN ref
) > 0
)
RETURN doc
)
LET base_scores = (
FOR doc IN filtered_docs
// Iterate over each cvss version
FOR cvss IN VALUES(doc.x_cvss)
LET base_score = cvss.base_score
FILTER base_score != null
RETURN base_score
)
LET mean_base_score = AVERAGE(base_scores)
LET median_base_score = MEDIAN(base_scores)
RETURN {
mean_base_score: mean_base_score,
median_base_score: median_base_score
}
[
{
"mean_base_score": 6.51316089491444,
"median_base_score": 6.5
}
]
How many vulnerabilities are published on a Friday?
Is it me, or are most vulnerabilities published on a Friday?
LET monday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 1
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET tuesday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 2
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET wednesday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 3
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET thursday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 4
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET friday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 5
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET saturday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 6
COLLECT WITH COUNT INTO count
RETURN count
)[0]
LET sunday_count = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "vulnerability"
AND DATE_DAYOFWEEK(doc.created) == 7
COLLECT WITH COUNT INTO count
RETURN count
)[0]
How wrong I was!
Most vulnerable products
Finally, I wanted to see the products with the most vulnerabilities discovered in them;
LET unique_vulnerable_criteria = (
FOR doc IN nvd_cve_vertex_collection
FILTER doc.type == "indicator"
// Extract the unique vulnerable criteria for each document
LET criteria = (
FOR cpe IN doc.x_cpes.vulnerable
LET parts = SPLIT(cpe.criteria, ":")
RETURN DISTINCT CONCAT(parts[3], ":", parts[4])
)
// Return the unique criteria for the document
RETURN DISTINCT FLATTEN(criteria)
)
FOR criteria_list IN unique_vulnerable_criteria
FOR criteria IN criteria_list
COLLECT vulnerable_criteria = criteria WITH COUNT INTO count
SORT count DESC
RETURN {
vulnerable_criteria: vulnerable_criteria,
count: count
}
UNIX OS’s Debian, Fedora, Ubuntu, OpenSUSE, and Redhat make up the top 5 products with most vulnerabilities published.
Keep in mind my search only considers vendor and product, it does not split the results by product version. Thus products that have multiple versions, and have likely been around for longer, are considered as one. Even so, given these are open-source projects, it is not surprising.
For me, the biggest takeaway from this data is the amount of firmware listed, specifically from Qualcomm, highlighting the ever-increasing risk of vulnerabilities at the edge (aka Internet of Things).
Vulmatch
Straightforward vulnerability management. Know when software you use is vulnerable, how it is being exploited, and how to detect an attack.
Discuss this post
Head on over to the DOGESEC community to discuss this post.
Never miss an update
Sign up to receive new articles in your inbox as they published.