138% more reported CVEs per day in 2026 than last year.New: always-on Fixboxes. Vulnerable dependencies, fixed and verified.

Vulnerability management

What is CWE?

CWE, the Common Weakness Enumeration, is a community-developed list of common software and hardware weaknesses: the kinds of mistakes, such as SQL injection or a use after free, that lead to vulnerabilities. A CVE names one vulnerability in one product. A CWE names the type of flaw behind it. Each entry has an ID, such as CWE-79 for cross-site scripting, so that advisories, scanners and training all describe the same weakness in the same way.

Updated

Who runs CWE

CWE is managed by the Homeland Security Systems Engineering and Development Institute (HSSEDI), operated by MITRE, and sponsored by the US Cybersecurity and Infrastructure Security Agency (CISA). The content comes from a community of vendors, researchers and government. MITRE updates the list three to four times a year. Version 4.21, released on October 28, 2026, has 984 weaknesses, 431 categories and 62 views. Everything is free at cwe.mitre.org.

How CWE is organized

CWE has three kinds of entries.

  • Weaknesses are the flaws themselves, such as CWE-89 SQL Injection.
  • Categories group weaknesses that share a characteristic, such as a type of error. A category is not a weakness and is never used to classify a vulnerability.
  • Views are ways of looking at the list for a purpose. The Complete Weakness Hierarchy (CWE-1000) arranges all weaknesses in a tree, the Software Development view (CWE-699) groups them the way developers think, and CWE-1003 holds the roughly 130 weaknesses most often seen in the National Vulnerability Database (NVD).

Weaknesses come at four levels of abstraction, from the most general to the most specific:

Level What it describes Example
Pillar The most abstract theme. The complete hierarchy starts with 10 of them CWE-707 Improper Neutralization
Class A weakness independent of any language or technology CWE-74 Injection, in output used by another component
Base Specific enough to say how to detect and prevent it CWE-89 SQL Injection
Variant Tied to a specific language or technology CWE-564 Hibernate query built from external input

Some entries combine weaknesses: a chain is a sequence in which one weakness creates the conditions for the next, and a composite needs several weaknesses present at once. MITRE's guidance is to classify a vulnerability at the Base or Variant level wherever possible, and to fall back to a Class only when nothing more specific fits.

How CWE IDs appear in CVE records

When a CVE Numbering Authority (CNA) publishes a CVE record, it can name the root cause as a CWE ID in the record's problem type. For the xz backdoor, CVE-2024-3094, the CNA recorded CWE-506 Embedded Malicious Code. NVD analysts may add their own CWE mapping to the same CVE, so a vulnerability can carry more than one. GitHub Security Advisories list CWE IDs too. The mapping is done by people, and it is not always precise: a record mapped to a broad Class, or to NVD's NVD-CWE-noinfo (insufficient information), tells you little about the root cause. See what a CVE is and what a security advisory contains.

CWE vs CVE vs CVSS vs CAPEC

These four are often confused, because they appear side by side in the same advisory.

Answers Example Run by
CWE What type of mistake is this? CWE-89 SQL Injection MITRE, sponsored by CISA
CVE Which specific vulnerability, in which product? CVE-2024-3094 in xz The CVE Program and its CNAs
CVSS How severe is it? A base score from 0.0 to 10.0 FIRST
CAPEC How do attackers exploit this kind of weakness? CAPEC-66 SQL Injection MITRE

The four connect: a CVE record names its CWE and carries a CVSS score, and each CAPEC attack pattern lists the CWE weaknesses it exploits (CAPEC-66 lists CWE-89, for example). None of them says how likely exploitation is; for that, look at EPSS and the CISA KEV catalog.

The CWE Top 25 Most Dangerous Software Weaknesses

Every year MITRE ranks the weaknesses behind the most frequent and most severe vulnerabilities. The 2025 list is based on 39,080 CVE records published between June 1, 2024 and June 1, 2025. A weakness's score combines how often it is mapped to a CVE in the NVD with the average CVSS base score of those CVEs. The top 10 of the 2025 CWE Top 25:

Rank CWE Weakness CVEs in KEV
1 CWE-79 Cross-site scripting 7
2 CWE-89 SQL injection 4
3 CWE-352 Cross-site request forgery (CSRF) 0
4 CWE-862 Missing authorization 0
5 CWE-787 Out-of-bounds write 12
6 CWE-22 Path traversal 10
7 CWE-416 Use after free 14
8 CWE-125 Out-of-bounds read 3
9 CWE-78 OS command injection 20
10 CWE-94 Code injection 7

The last column counts the CVEs in the list's dataset that are also in CISA's Known Exploited Vulnerabilities catalog. It shows why rank and real-world risk differ: cross-site scripting tops the list by volume, while OS command injection, at rank 9, had the most exploited CVEs. Positions 11 to 25 include classic buffer overflow (CWE-120), unrestricted file upload (CWE-434), deserialization of untrusted data (CWE-502), missing authentication for a critical function (CWE-306) and server-side request forgery (CWE-918).

CWE in the software supply chain

Most CWEs describe mistakes in your own code. A few describe the risks of the code you take from others:

CWE Weakness In practice
CWE-1395 Dependency on Vulnerable Third-Party Component Shipping a library with a known CVE
CWE-1104 Use of Unmaintained Third Party Components Depending on a package nobody patches any more
CWE-829 Inclusion of Functionality from Untrusted Control Sphere Loading code from a source you do not control
CWE-506 Embedded Malicious Code A backdoored release, such as xz in 2024
CWE-1357 Reliance on Insufficiently Trustworthy Component Building on a component that does not meet your security or maintenance needs

These weaknesses are not found by reading your own source. You find them by knowing exactly which third-party packages went into each build. See what is a supply chain attack.

How to use CWE

  • Fix root causes, not just CVEs. Group your own past vulnerabilities by CWE. If the same weakness keeps coming back, a framework change, a safe API or a lint rule removes the whole class instead of one instance.
  • Secure by design. CISA's Secure by Design alerts ask vendors to eliminate whole classes of vulnerability, such as cross-site scripting, OS command injection and directory traversal, all of them in the Top 25.
  • Map tools to a shared language. Static analysis (SAST) rules and penetration test findings are commonly tagged with CWE IDs, so results from different tools can be compared and deduplicated. See SAST vs SCA.
  • Write requirements and train. Product security requirements and developer training can name the CWEs that matter for your stack, such as the Top 25 entries for your languages.
  • Classify your own advisories. When you publish a vulnerability in your product, give the CWE so customers know what kind of flaw it was. See coordinated vulnerability disclosure.

CWE does not tell you what to fix first. Severity, exploitation and whether the code is in your product do. See vulnerability prioritization.

Where CRACI fits

CWE-1395 is the weakness CRACI addresses directly: shipping a component with a known vulnerability. CRACI records the packages each GitHub Actions job actually fetched, including transitive dependencies, and each package shows its advisories (CVE, GHSA and other IDs) with their severity and the packages that pull it in. Build history shows every past build that contained a vulnerable dependency, so you know which releases carry it, and security teams triage the findings and send them to the teams that own the fix. A known malicious release with a published CVE, the CWE-506 case, is flagged like any other vulnerability. See vulnerability tracking, or our vulnerability analyses.

When CRACI is not the answer

  • Weaknesses in your own code. Finding CWE-79 or CWE-89 in code you wrote is the job of SAST, code review and testing. CRACI does not scan source code, by design, and works alongside the SAST tool you use.
  • New malicious packages. CRACI does not detect malware by behavior or reputation. Until an advisory exists, its egress policy limits what such a package can send out of the build.

CWE: frequently asked questions

What is CWE in simple terms?

A numbered list of the kinds of mistakes that cause vulnerabilities, such as SQL injection or a buffer overflow. It gives every weakness type a shared ID, so tools, advisories and people use the same name for it.

What is the difference between CWE and CVE?

A CWE is a type of weakness, such as CWE-89 SQL Injection. A CVE is one specific vulnerability in one product, such as CVE-2024-3094 in xz. A CVE record usually names the CWE behind it.

What is the CWE Top 25?

MITRE's yearly ranking of the most dangerous software weaknesses, based on how often each is mapped to CVEs and how severe those CVEs are. The 2025 list is led by cross-site scripting (CWE-79), SQL injection (CWE-89) and CSRF (CWE-352).

Who assigns a CWE to a CVE?

The CVE Numbering Authority that publishes the CVE record can name a CWE in it. NVD analysts may add their own mapping, so the same CVE can show more than one.

How many CWEs are there?

CWE version 4.21, released on October 28, 2026, lists 984 weaknesses, plus 431 categories and 62 views. MITRE updates the list three to four times a year.

Is CWE the same as the OWASP Top 10?

No. The OWASP Top 10 is a short list of web application risk categories, and each category maps to several CWEs. CWE is the full catalog of weakness types those categories are built from.

Know which builds carry a vulnerable component

Record every package your GitHub Actions jobs fetch, and see the advisories and severity for each one.

Book a demo