What is a CVE?
A CVE (Common Vulnerabilities and Exposures) identifier is a public label for one specific, disclosed vulnerability, written as CVE-year-number. It lets everyone talk about the same flaw under the same name. It does not say how bad the flaw is, and it is not a fix.
Quick facts
- Format
CVE-YYYY-NNNN: the prefix, a four-digit year, and a sequence number of at least four digits (CVE FAQs)- Program
- The CVE Program, with CVE Numbering Authorities (CNAs) assigning IDs (CVE Program overview)
- Identifies
- One publicly disclosed vulnerability, so that tools, advisories and people refer to the same thing
- Core record
- An ID, a description, affected products and references (CVE process)
- Not in the identifier
- A severity rating, a patch, or an assessment of risk to your own systems
- Enrichment
- The National Vulnerability Database (run by NIST) works from CVE records and adds analysis such as CVSS scores, CWE weakness types and product names
- Example on this site
- Log4Shell is CVE-2021-44228; see the NVD entry for CVE-2021-44228
The ID format
A CVE ID has three parts: the letters CVE, a year, and a sequence number. In CVE-2021-44228 the year is 2021 and 44228 is a sequence number; it does not count how many flaws were found that year. The sequence has at least four digits and grows longer when a year needs more numbers, so do not write a pattern that expects exactly four.
The year is a poor guide to dates. It reflects when the ID was reserved or assigned, which can be earlier or later than the day the flaw was found or the record published. Numbers are not consecutive in time either, because numbering authorities receive blocks of them and use them at their own pace.
One ID covers one vulnerability. When a single product ships several separate flaws, each gets its own ID; when several products carry the same flaw through a shared library, they usually point to the library's ID. That second case is why a flaw with no fix yet and a flaw with a published ID are different questions: an ID can exist before, at, or after the moment a fix does.
Who assigns CVE IDs
The CVE Program does not hand out IDs from one desk. It is organised in tiers, described in the CVE Program overview. CVE Numbering Authorities (CNAs) are organisations authorised to assign IDs within a defined scope: a software vendor for its own products, a research or coordination body for a wider area. Roots oversee groups of CNAs. MITRE is one of the program’s roots and acts as a CNA of last resort for products that no other CNA covers.
Most people who find a flaw therefore report it to the vendor, and the vendor, as CNA, requests the ID. If the product has no CNA, the report can go to a CNA of last resort. The practical effect is that the quality of a record depends on who wrote it: a vendor CNA tends to know its own version numbers, and a catch-all authority may know less. The CVE process page describes the steps from reservation to publication.
What a record holds, and what it does not
A published CVE record is deliberately small. It carries the ID, a short description of the flaw, the products and versions affected as the CNA understands them, and references: links to advisories, bug reports or fixed releases. Treat the references as the useful part, because that is where a vendor says which version to move to.
Three things people expect are not part of that core.
- A severity rating. Scores come from elsewhere. A CNA may add a CVSS score to its record, and NVD may add its own. Which version of CVSS is used differs between records; CVSS explains the scale and why a base score is not a statement about your exposure.
- A patch. The record describes and points. The remedy is a release or a configuration change from the vendor.
- Proof that you are affected. Whether your deployment is exposed depends on which version you run, how it is configured and whether the vulnerable code path is reachable. The record lists what the CNA knew when it wrote it.
Records are also revised after publication: descriptions are corrected, references added, affected ranges widened. An ID you saw in an advisory last year may read differently today.
NVD and CWE
The National Vulnerability Database is a separate resource, maintained by the US National Institute of Standards and Technology. It takes CVE records and adds analysis: severity scores, product identifiers and weakness classification. When a page says "NVD rates it 10.0", as the Log4Shell page does for CVE-2021-44228, the score is NVD's work on top of the CVE entry, not part of the identifier itself. The fact that NVD analysis can lag behind a CVE's publication is one reason a fresh record sometimes appears without a score.
CWE (Common Weakness Enumeration), maintained by MITRE, is a catalogue of weakness types rather than individual flaws. The relationship is class and instance: a CWE entry names a kind of mistake, such as improper input validation (CWE-20), and a CVE is one real case of a mistake in one product. NVD lists CWE numbers for CVE-2021-44228, among them CWE-20 and CWE-502. The OWASP supply chain category sits a level above both, as a way of grouping the problem by where it comes from.
Using CVE IDs with your dependencies
The identifier earns its keep as a join key. If you know which components and versions you run, an advisory that names a CVE becomes a quick lookup rather than a research task. The work is on your side:
Keep a list of what you run, including libraries pulled in by other software; an ID cannot match what you have never written down. Match on the product and version ranges in the record, then read the vendor's reference, not only the summary. Rank by your own exposure. A high score on a component that is not reachable from outside may matter less than a modest one on your login form. Record the decision, including "not affected, because", so the same question is not reopened at every scan.
Scanners that report CVE IDs are only as good as the inventory and the data behind them, and they report that a version is in range, not that the flaw is exploitable in your setup. This is the daily reality behind the "dependencies you did not choose" theme of the supply chain failures category.
Where a WAF helps and where it does not
A web application firewall reads HTTP requests. It has no concept of a CVE ID; what it can have is a rule that recognises the shape of requests which try to use a particular flaw, such as a pattern in a header or parameter. If such a rule exists and the attack arrives over HTTP through the filter, the request can be flagged or refused. That is the idea behind virtual patching.
The limits are plain. The filter does not repair the vulnerable code. It cannot see input that reaches the component by another path, and a rule written for a known exploit shape can miss a variant. Not every CVE is reachable over HTTP at all. Updating the software stays the remedy.
Bridge WAF has three modes: off, monitor and block. For a newly disclosed flaw, start in monitor. Matches are recorded and every request is still served, so the activity log shows what would be refused, and you can check for false positives before you move to block. Read monitor mode first if you have not used it.
Questions
What does CVE stand for?
CVE stands for Common Vulnerabilities and Exposures. It is the name of the identifier scheme and of the program that coordinates it. A CVE ID such as CVE-2021-44228 points to one publicly disclosed vulnerability.
Does a CVE tell me how severe a vulnerability is?
No. The identifier and the core record name and describe the flaw and list references. Severity scores such as CVSS come from the CNA that published the record or from NVD, and a score describes the flaw in general, not the risk to your own systems.
Is there a patch in the CVE record?
Not as such. A record may list references to vendor advisories that point to fixed versions, but the CVE entry itself is a description and a set of links. The fix is whatever the vendor of the affected software ships.
Does the year in a CVE ID mean the flaw was found that year?
Not necessarily. The year reflects when the ID was reserved or assigned, which can differ from when the flaw was discovered or when the record was published.
Can a WAF block a CVE?
A WAF does not block identifiers. It can recognise the shape of requests that try to use a specific flaw and flag or refuse them, which lowers exposure while you update the affected software. It does not remove the flaw. Start in monitor mode and read the activity log before switching to block.