Score a vulnerability in CVSS 4.0 or 3.1, and see why the same finding gets a different number in each.
Choose a value for each metric, or paste a CVSS 4.0 or 3.1 vector into the Vector box. The score updates either way.
CVSS rates how severe a vulnerability is on a scale from 0.0 to 10.0. FIRST, the organization that owns the standard, maps the number to five bands, and the bands are the same in 3.1 and 4.0:
A score is only half a finding. Three habits keep it useful in a report:
Nobody can check a bare 7.1. They can check the vector it came from. FIRST's 4.0 specification asks anyone who publishes CVSS data to give "both the score and the vector string".
In FIRST's words, "CVSS-B Base scores are not risk, and should not be used alone for patch prioritization." Your client's exposure and whether an exploit is out there decide the risk.
FIRST names each combination: CVSS-B for Base Metrics only, CVSS-BT with Threat, CVSS-BE with Environmental, and CVSS-BTE with all three.
FIRST's CVSS v4.0 Examples document scores real CVEs in both versions, with both vectors. We took the 36 of its 42 pairs where FIRST gives one Base score on each side (the note under the full table says which 6 we left out), and scored them again with the same code this calculator uses. Every number matched FIRST's. Not one score stayed the same: 20 went up, 16 went down, and 12 moved to a different severity band.
| Finding | CVE | CVSS 3.1 | CVSS 4.0 | Severity band |
|---|---|---|---|---|
| NGINX mp4 module memory corruption | CVE-2022-41741 | 7.0 High | 7.3 High | Same band |
| Cisco Firepower sftunnel registration hash exposure | CVE-2020-3549 | 8.1 High | 7.7 High | Same band |
| Crypto modules not FIPS-validated in FIPS mode (Red Hat OpenShift) | CVE-2023-3089 | 7.5 High | 8.3 High | Same band |
| Misleading warning lets a user bypass a security feature (Acrobat Reader DC) | CVE-2021-44714 | 3.3 Low | 4.6 Medium | Up to Medium |
| RocketChat LiveChat blind self XSS | CVE-2022-21830 | 6.1 Medium | 5.1 Medium | Same band |
| Junos OS EX4650 improper packet forwarding | CVE-2022-22186 | 7.2 High | 6.9 Medium | Down to Medium |
| Oracle VM VirtualBox flaw (high-privileged local attacker) | CVE-2023-21989 | 6.0 Medium | 5.9 Medium | Same band |
| VMware Workstation and Fusion use-after-free | CVE-2020-3947 | 9.3 Critical | 9.4 Critical | Same band |
| Unauthenticated changes via physical access (Becton Dickinson infusion PCU) | CVE-2023-30560 | 6.8 Medium | 8.3 High | Up to High |
| Heartbleed (OpenSSL) | CVE-2014-0160 | 7.5 High | 8.7 High | Same band |
| Log4Shell (Apache log4j) | CVE-2021-44228 | 10.0 Critical | 9.3 Critical | Same band |
| Shellshock (Bash) | CVE-2014-6271 | 9.8 Critical | 9.3 Critical | Same band |
| Juniper proxy ARP denial of service | CVE-2013-6014 | 9.3 Critical | 6.4 Medium | Down to Medium |
| SQL injection reads server files (PrestaShop) | CVE-2023-30545 | 6.5 Medium | 7.1 High | Up to High |
| Credentials sent in clear text over HTTP (Bosch firmware) | CVE-2021-23846 | 5.9 Medium | 8.2 High | Up to High |
| SIP memory leak denial of service (Junos OS) | CVE-2023-22394 | 7.5 High | 8.7 High | Same band |
| Reflected cross-site scripting (Zimbra) | CVE-2022-24682 | 6.1 Medium | 5.1 Medium | Same band |
| Stored cross-site scripting (SharePoint) | CVE-2020-0926 | 5.4 Medium | 5.1 Medium | Same band |
| Cross-site scripting with impact beyond the user's browser (Dolibarr) | CVE-2024-55228 | 9.0 Critical | 8.5 High | Down to High |
| Cross-site request forgery (WordPress plugin) | CVE-2023-5602 | 4.3 Medium | 5.1 Medium | Same band |
| Privilege escalation (Cisco ASA and FTD remote access VPN) | CVE-2022-20759 | 8.8 High | 7.7 High | Same band |
| Privilege escalation to root (Cisco IOS XE SD-WAN CLI) | CVE-2021-34724 | 6.0 Medium | 8.3 High | Up to High |
| Remote code execution (Microsoft Word) | CVE-2023-28311 | 7.8 High | 8.5 High | Same band |
| Spring4Shell | CVE-2022-22965 | 9.8 Critical | 9.2 Critical | Same band |
| Secure boot bypass via physical access (Cisco) | CVE-2022-20826 | 6.4 Medium | 5.4 Medium | Same band |
| Information disclosure (Oracle E-Business Suite) | CVE-2022-21500 | 7.5 High | 8.7 High | Same band |
| Information disclosure (Ericsson ENM logs) | CVE-2021-32570 | 4.9 Medium | 6.9 Medium | Same band |
| Command injection (Confluence) | CVE-2022-26134 | 9.8 Critical | 9.3 Critical | Same band |
| ACL bypass (Cisco ASA) | CVE-2023-20245 | 5.8 Medium | 6.9 Medium | Same band |
| Server-side request forgery (JBoss EAP) | CVE-2024-1233 | 7.3 High | 6.9 Medium | Down to Medium |
| Out-of-bounds write (CODESYS, operational technology) | CVE-2022-47379 | 8.8 High | 9.4 Critical | Up to Critical |
| Improper authentication and authorization on insulin pump radio link (Insulet Omnipod) | CVE-2020-10627 | 8.1 High | 8.6 High | Same band |
| Unbounded recursion (MIT Kerberos 5) | CVE-2020-28196 | 7.5 High | 8.7 High | Same band |
| Unauthorized configuration commands (Cisco Firepower Management Center) | CVE-2023-20048 | 9.9 Critical | 6.4 Medium | Down to Medium |
| Arbitrary file read (Jenkins CLI) | CVE-2024-23897 | 9.8 Critical | 10.0 Critical | Same band |
| Windows Kerberos denial of service | CVE-2026-42903 | 6.5 Medium | 7.1 High | Up to High |
Base score against Base score. FIRST's document has 42 pairs and we left out 6: five where it prints two different 4.0 scores for the same CVE, and regreSSHion (CVE-2024-6387), whose 4.0 vector lowers Exploit Maturity. Four of FIRST's 4.0 vectors set Exploit Maturity to Attacked, which scores the same as leaving it out.
Most of the moves follow one of three patterns:
9 of the 11 that 3.1 scored as Scope Changed went down. 3.1 gives Scope Changed a boost, and 4.0 has no Scope at all. Log4Shell lost that boost, from 10.0 to 9.3, with all its damage on the vulnerable system. Where FIRST recorded the damage only on other systems, 6 of 7 fell too, some by a lot: reflected XSS went from 6.1 to 5.1, and CVE-2023-20048, a Cisco management console flaw, from 9.9 Critical to 6.4 Medium. The one that rose was a Cisco ASA ACL bypass, 5.8 to 6.9.
All 9 with Scope Unchanged and one High impact (confidentiality only, or availability only) scored higher in 4.0. Heartbleed went from 7.5 to 8.7, and the SQL injection example crossed from 6.5 Medium to 7.1 High. Credentials sent in clear text over HTTP jumped from 5.9 to 8.2. The attacker has to be on the network path: 3.1 scored that as High attack complexity, 4.0 as Low complexity with Attack Requirements Present.
All 5 network findings rated High on C, I and A, with Scope Unchanged and no damage beyond the vulnerable system, went down. Where the metrics match in both versions, as for Confluence and Shellshock, 9.8 became 9.3. The other three also picked up Attack Requirements Present in FIRST's 4.0 vector. Add damage to other systems and the same kind of finding goes up instead: Jenkins went from 9.8 to 10.0, and CODESYS from 8.8 High to 9.4 Critical.
Not every move comes from the new math. For some findings FIRST's analysts also recorded something different in the 4.0 vector, like that Attack Requirements value. And FIRST picked these examples to show what 4.0 changes, so read the patterns as a reason to check, not as rules.
Moving a report to 4.0 means a new 4.0 vector for every finding: a 3.1 vector won't even score in 4.0. Check the three patterns above first. If your client sets fix deadlines by severity, a band change moves the deadline. Mixing both versions in one report is its own problem. Our guide to CVSS 4.0 for pentest reports covers it.
A vector records the metric values behind a score. It starts with CVSS: and the version, then lists each metric as a code and a value, separated by slashes. Every Base metric has to be there. The rest are optional and count as Not Defined when you leave them out.
CVSS:3.1Version
AV:NNetwork
AC:LLow complexity
PR:LLow privileges
UI:NNo user
S:UUnchanged
C:HHigh
I:NNone
A:NNone
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N says: over the network, nothing hard about it, needs a low-privileged account, no user involved, the damage stays in the vulnerable component, and the attacker can read data they shouldn't but change nothing. That's the SQL injection above, scored 6.5.| Code | Metric | CVSS 3.1 values | CVSS 4.0 values |
|---|---|---|---|
AV | Attack Vector | N Network, A Adjacent, L Local, P Physical | Same |
AC | Attack Complexity | L Low, H High | Same |
AT | Attack Requirements | Not in 3.1 | N None, P Present |
PR | Privileges Required | N None, L Low, H High | Same |
UI | User Interaction | N None, R Required | N None, P Passive, A Active |
S | Scope | U Unchanged, C Changed | Retired |
C, I, A | Confidentiality, Integrity, Availability | H High, L Low, N None | VC, VI, VA for the vulnerable system and SC, SI, SA for subsequent systems: H, L, N |
In 4.0, User Interaction splits Required into Passive and Active. FIRST's 4.0 specification puts most stored XSS under Passive, where the user only has to load a page the attacker changed, and reflected XSS under Active, where the user has to submit the attacker's string themselves.
Beyond Base, 3.1 has optional Temporal and Environmental Metrics. 4.0 renamed Temporal to Threat and kept only Exploit Maturity, which counts as Attacked, the worst case, when you leave it unset. 4.0 also adds Supplemental metrics, such as Safety and Automatable, which never change the score.
In 3.1 it's arithmetic, set out in FIRST's 3.1 specification.
With Scope Unchanged, the impact sub-score is 1 - (1 - C) x (1 - I) x (1 - A), Impact is 6.42 times that, and Exploitability is 8.22 x AV x AC x PR x UI. The Base score is Impact plus Exploitability, capped at 10 and rounded up to one decimal. Scope Changed uses a different Impact formula and multiplies the sum by 1.08. For the SQL injection vector: Impact is 6.42 x 0.56 = 3.5952, Exploitability is 8.22 x 0.85 x 0.77 x 0.62 x 0.85 = 2.8353, the sum is 6.4305, and that rounds up to 6.5.
4.0 has no formula like 3.1's. FIRST's experts ranked groups of similar vectors, called MacroVectors, and a vector's score comes from a lookup table plus interpolation inside its group: about 15 million vectors map onto 101 possible scores. That's why the moves in the chart above don't follow one metric weight. This calculator runs FIRST's reference code for 3.1 and, for 4.0, the same code Dradis runs on every issue.
We put this calculator on every issue in Dradis CE and Dradis Pro, for CVSS 3.0, 3.1 and 4.0. Pick the version, set the metrics, and Update Issue writes the vector, score and severity into the issue's fields (CVSSv4.BaseVector, CVSSv4.BaseScore and so on, or CVSSv3.Vector and CVSSv3.BaseScore for 3.1), where your report template picks them up. The field output box above uses the same field names, so a score you work out here pastes straight into an issue.
On a team, the version and the vector live on each issue, so whoever reviews the report can see which version every tester used before it goes to the client. Score a finding on a real issue in the Dradis CE sandbox (it resets every 20 minutes), or read how it works in the CVSS calculator guide.
The current version is 4.0, which FIRST released on 1 November 2023 with new metrics and a new way of working out the score. The version before it, 3.1, clarified 3.0 without adding metrics, so a 3.0 vector uses the same metrics as a 3.1 one.
It's the top of the scale, inside the Critical band (9.0 to 10.0). In 3.1, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H scores 10.0. In 4.0 it takes high impact on the vulnerable system and on subsequent systems: FIRST's Jenkins example, CVE-2024-23897, scores 10.0 that way.
Attack Vector: Physical. The attacker has to touch or manipulate the vulnerable device, even briefly, as in an evil maid attack. In 3.1 it carries the lowest weight of the four Attack Vector values.
Because 4.0 is a different scoring system, not a new edition of 3.1, so there's no converting a 3.1 number: score the finding again in 4.0. The biggest drops come on findings 3.1 scored as Scope Changed, and some findings go the other way. In FIRST's examples, a Juniper proxy ARP denial of service fell from 9.3 Critical to 6.4 Medium, while a Cisco IOS XE privilege escalation to root (CVE-2021-34724) rose from 6.0 Medium to 8.3 High.
No. FIRST says CVSS measures severity, not risk, so a Base score alone shouldn't set the order you patch in. Threat and Environmental Metrics, or your client's own context, bring it closer to risk.
Loading form...
Your email is kept private. We don't do the spam thing.