Skip to content

Assess findings with VEX

A finding is a vulnerability that Affectra matched to a release. The list has one row per CVE, however many advisories report it: a finding group. A VEX statement is your decision about one finding: is this release affected or not, and why. You record statements so that your customers and authorities can see what you did about each vulnerability. You can export them as OpenVEX, CycloneDX or CSAF.

Everyone in your organization can read findings and statements. To record a statement you need the member role or higher. See Members and roles.

Open Releases, select the product and the release, and stay on the Vulnerabilities tab. Findings appear once Affectra has matched the release’s components, usually within a minute of an upload.

The line above the list says how current the results are. For example: Checked against vulnerability data 2 hours ago. Select Details to see when each source was checked and when it last changed. The release warns you when a source is stale, or was never loaded, because results may then be incomplete. It also says when components are still waiting to be matched, and how many components could be checked at all. Vulnerability sources lists the sources and how often they refresh.

Each row is one CVE. A CVE that the NVD, OSV, a distribution and a vendor all publish is one row, not four. An advisory without a CVE is a row of its own, with its own ID in place of the CVE, unless it is an alias of another advisory without a CVE, such as a Go advisory (GO-…) that names the GitHub advisory (GHSA-…) for the same problem. Those are one row, shown with the smaller of the two IDs. A row shows:

  • the CVE, in monospace, and the title. NVD’s title is used when the NVD has the CVE, else the title of the first advisory;
  • the sources: a strip with one chip for each advisory behind the row. See The source strip;
  • Known exploited, when a CVE of the finding is in CISA’s Known Exploited Vulnerabilities catalog or on ENISA’s EU KEV list. These vulnerabilities are being used in attacks;
  • the severity and the CVSS score, for example high 7.5. They come from the NVD. Without NVD data there is no score, and the line above the list says CVSS scores are unavailable: NVD sync is off;
  • EPSS, the probability of exploitation in the next 30 days. From 10 % it is highlighted;
  • the EUVD IDs of its CVEs, such as EUVD-2025-4893: the IDs in ENISA’s EU Vulnerability Database, which EU customers and CRA reports cite. Each links to its page in the EUVD;
  • Vendor advisories: the CSAF advisories of vendors and coordinators, such as Siemens or BSI, that name its CVEs, each linking to the advisory, with the versions that fix the products they list. See Vendor advisories;
  • the VEX status, if you recorded a statement;
  • Sources disagree and Statements differ, when they apply. See below;
  • New, when the finding was first found less than 7 days ago;
  • the components of the release that have it, from all of the advisories. A component found through its CPE is marked CPE, and one found through your own package mapping is marked your mapping.

Findings are ordered known exploited first, then by CVSS score, then by EPSS.

The list shows the Open findings. Select All to include the ones a statement resolved. The findings come a page at a time, the most urgent first: select Load more under the list for the next ones. Open (n), All (n) and the number on the tab count CVEs, not advisories, and appear once every finding is loaded. Use the search box to find a CVE, an advisory ID, an EUVD ID or a component among the findings loaded so far; while more are waiting, the list says so. When you search for an advisory’s ID, the row that has it stays and the advisory’s chip is highlighted.

Under the title, each advisory behind the row has a chip, in this order: the NVD, then OSV advisories (GitHub, Debian and others), then vendors’ and coordinators’ CSAF advisories. Each chip links to the advisory’s page where there is one.

  • A filled dot (●) means the advisory matched a component of this release.
  • A hollow dot (○) means the advisory names a component of this release but says that its version is not affected, for example because it is fixed in an earlier version.
  • A small mark after the ID shows that you made a statement on that advisory itself.

Hover over or focus a chip to see the advisory’s title and the version that fixes it, where the data has one. A screen reader reads the same as one sentence, for example GHSA-…: not affected, fixed in 2.17.0. The dots differ in shape, not only in colour.

Sources disagree appears when one advisory says that a component of your release is affected and another advisory for the same CVE says that its version is not. Neither is wrong by itself: sources update at different times, and a vendor often knows better than a database. Select Details to see which source is the odd one out. The row stays open until you decide: only a statement resolves it, the disagreement does not.

Statements differ appears when you made a statement on one of the advisories, and it is not the same as the statement on the CVE. The statement on the CVE decides. See Statements on advisories and on CVEs.

Select Details on a row to see the evidence:

  • The Finding ID, such as Gateway/rev-b@1.4.0:CVE-2024-1234, with a button that copies it and Copy link, which copies a link that opens this finding. Use it in tickets, in chat and with your assistant. See Finding IDs.
  • A table with the components of the release as rows and the advisories as columns. Each cell says, in words, whether the advisory finds the component affected (●), says it is fixed in a version (○), or has no data on it (—). A row where one advisory says affected and another says fixed is tinted: that is the disagreement.
  • The Statements on the CVE and on each advisory, so you can see which one decides.
  • The Timeline, below.

Select Timeline of to choose which advisory’s timeline to show when the row has more than one. It starts with the best-ranked advisory.

The timeline in Details shows how the finding came about: when it was published, when it reached Affectra, when your SBOM was uploaded, when it was matched to this release, when it became known exploited, and your statements on it. It also lists the finding’s EUVD IDs. Under Why it’s here it says whether the finding was found when you uploaded the SBOM, came with a new or updated advisory, or came through your mapping. For the rules behind a finding, read Why a finding is or isn’t shown.

On a finding, select Assess. A panel titled Assess a vulnerability opens with the CVE filled in, and a line that says which advisories the statement covers, for example Covers NVD, GHSA-… and DSA-… for this release. You can also open the VEX tab and select New statement to make a statement about any CVE or advisory.

  1. In Vulnerability, check the ID. Type one if the field is empty, for example CVE-2023-0286 or DSA-5343-1.
  2. Choose a Status: Not affected, Affected, Fixed or Under investigation.
  3. Fill in what the status needs, below.
  4. Select Record statement.
Status What you give The finding
Not affected A Justification, an Impact statement, or both. You need at least one. Resolved
Affected An Action statement: what should users do? It is required. Stays open
Fixed Nothing more. The vulnerability is fixed in this release. Resolved
Under investigation Nothing more. It records that you are looking into it. Stays open
  • A justification is one of five values from the OpenVEX specification. The picker shows each value with what it means. VEX statuses and justifications lists them.
  • Only Not affected takes a justification. The panel shows the field for that status only, and the API refuses a justification on any other.
  • Impact statement is free text on why the release is not affected. Action statement is free text on what users should do, for example update to release 1.1. Each is at most 4000 characters.
  • If you leave the picker on None, explained below, the impact statement has to explain.

The panel tells you what is missing, and the API checks the same rules again.

Assess on a row records a statement on its CVE. It decides for every advisory in the row, and for every component. The statement on the CVE decides over statements on the single advisories. A statement on an advisory still counts when there is none on the CVE.

Some advisories, such as a Debian advisory, cover several CVEs. A statement on the advisory decides for the whole advisory. To assess its CVEs one by one, select one of the CVE IDs the panel offers.

When you have made no statement on the advisory itself, the statements on its CVEs decide:

  • If any CVE is Affected, the finding is Affected. Otherwise, if any is Under investigation, it is Under investigation.
  • When every CVE of the advisory has a statement, the finding is Fixed if any CVE is fixed, and Not affected otherwise.
  • When only some CVEs have a statement, the finding has no status and stays open.

Each CVE counts once. In a release, the tab’s number, Open and All count CVEs. On the overview, To assess, Known exploited and Resolved by VEX count each CVE once across all of your releases, so a CVE that is in five releases counts as one. A CVE is open on the overview when it is open in at least one release, and resolved when it is resolved in all of the releases that have it. The hint under To assess says how many releases the open findings are on, for example on 7 releases. A release’s own number, in the list of releases and under Most exposed releases, counts its CVEs.

To assess is the number to work down: the open findings with no statement or under investigation, which still need your assessment. A finding with the status Affected is assessed: its action statement tells your customers what to do, for example to update to a release with the fix. A release doesn’t change, so an affected finding stays open in that release for good: the release is vulnerable, and the fix ships in another release. The overview’s Units to upgrade counts the units that still run such a release; see See which units to upgrade. Affected findings don’t count under To assess; when there are any, its hint says how many, for example 1 assessed as affected. On the overview, a CVE counts as affected when it is affected in every release where it is open. The list of releases and Most exposed releases split each release’s open findings the same way, for example 8 to assess · 1 affected, or 1 affected once every open finding is assessed. Most exposed releases lists the releases of products that are still supported first, and the ones past their support end last. Within each group it puts the known exploited ones first, then the ones with the most findings still to assess.

A statement is never edited or deleted. To change your mind, record a new statement for the same vulnerability. The new one becomes the current statement, and the old one stays in the history.

On the VEX tab, Current shows the latest statement for each vulnerability, sorted by vulnerability ID. History shows every statement, newest first, including the ones that were replaced. Each row shows who made the statement and when.

A statement belongs to one release. A new release starts without statements, even when the same vulnerability appears in it.

When the same vulnerable component is reached the same way in another release, a statement you made on one release often holds for the other. Affectra does not copy statements by itself. An agent can propose to carry them over, and you decide.

These proposals appear under Proposals, in a list titled Carry-overs to review. Each row shows the proposed statement next to the earlier one. Select the rows you agree with and select Accept selected. Accepting a selection accepts all of it or none of it.

A carry-over rests on the earlier statement and on that release’s SBOMs. When someone records a newer statement on the earlier release for the same vulnerability, or uploads an SBOM to it, the carry-over becomes stale and can no longer be accepted.

Connect your own AI assistant to Affectra. It reads your data through Affectra, finds statements you made on other releases, and proposes VEX statements. It cannot make one. A person decides. Built for AI assistants explains why. An assistant that works in your repository can also check your code: Assess findings with your source code.

Proposals appear under Proposals, and on the release’s Proposals tab, whose number counts that release’s pending proposals. Each one shows:

  • The claim: what the agent proposes, next to the earlier statement if there is one;
  • Evidence path: the parts of your data that the agent cites. Affectra checked that they exist when the proposal was filed;
  • The agent’s explanation, as the agent wrote it. Check it against the evidence before you rely on it.

You then choose:

  • Accept writes the statement as proposed.
  • Accept with edits opens the statement for you to change, with an optional note on what you changed and why.
  • Reject needs a note on why. The agent reads it, so that it does not propose the same again.

An accepted statement is an ordinary statement. On the VEX tab it has a link from an agent proposal. A member can decide a VEX proposal. An API key cannot.

An agent never decides its own proposals: a person does. By default, that can be the person the agent works for. When a second person should check every proposal, an admin turns on Four eyes for agent proposals under Settings. Then you cannot decide a proposal filed with your own credentials, such as an assistant signed in as you, nor one from an API key that you created.

A proposal can no longer be decided when the release’s data changed after it was filed, for example a new SBOM or a statement written directly. It then shows as Outdated. A newer proposal for the same finding replaces an older one, which then shows as Superseded.

CI or your own tools can record statements too. The key needs the access Can propose changes, because a Read only key cannot write. See Upload SBOMs from other CI for creating a key. $release is the release’s ID.

Terminal window
curl --fail-with-body --silent --show-error \
--request POST \
--header "Authorization: Bearer $AFFECTRA_API_KEY" \
--header "Content-Type: application/json" \
--data '{"vulnerability":"CVE-2023-0286","status":"not_affected","justification":"vulnerable_code_not_in_execute_path"}' \
"https://affectra.example.com/v1/orgs/acme/releases/$release/vex"

GET .../vex lists the current statements, and GET .../vex?history=true lists all of them. Another value than true or false is refused with 400. A statement that breaks a rule above is refused with 400, and the message says which rule.