What Is a CVE? How to Read One and Patch Fast

Devops & Infrastructure, Security, and What Is

Every few weeks a security headline lands with a string like CVE-2026-109618 in it, and someone on the team asks the same question: does this affect us, and how fast do we need to move?

That question is easier to answer once you know what a CVE actually is, what the numbers attached to it mean, and why the window between disclosed and being exploited has been shrinking. This guide covers all three, then gives you a patching routine that works for teams who deploy to their own servers.

What is a CVE?

CVE stands for Common Vulnerabilities and Exposures. It is a public catalogue of known security flaws in software and hardware, where each flaw gets a unique identifier such as CVE-2026-109618. The identifier format is CVE-<year>-<number>: the year the ID was assigned or the flaw was published, then a sequence number.

The CVE Program was started by MITRE in 1999 and is sponsored by the US Cybersecurity and Infrastructure Security Agency (CISA). Its job is narrow: give every publicly disclosed vulnerability one agreed name, so that a vendor advisory, a scanner report, a patch note and a news article are all talking about the same bug.

A CVE record does not tell you how bad the flaw is or whether anyone is exploiting it. It is a label plus a short description and references. The rest of the picture comes from other systems.

CVE, CVSS, NVD and KEV: who does what

These four acronyms get mixed up constantly. They answer different questions:

Acronym What it is Question it answers
CVE The identifier and description of a flaw Which vulnerability are we talking about?
CVSS A 0–10 severity score maintained by FIRST How bad could this be in theory?
NVD The US National Vulnerability Database (NIST), which enriches CVE records with scores and affected-product data Which versions are affected, and what is the score?
KEV CISA's Known Exploited Vulnerabilities catalogue Is anyone actually exploiting this right now?

The last row matters most in practice. A CVSS score measures potential impact. Inclusion in the CISA KEV catalogue means there is evidence of real-world exploitation. A 9.8 in a component you never expose to the internet is a scheduled upgrade. A 7.5 that is on the KEV list and affects your public web server is a today problem.

How to read a CVE entry

When an advisory lands, check these five things in order:

  1. Do we run the affected product at all? Check the affected-versions list against what is actually installed, not what you think is installed.
  2. Is a fixed version available? The record or the vendor advisory names the patched release.
  3. Is it exposed? A remote code execution flaw in a service listening on the public internet is a different animal from one that needs local access.
  4. Is it on the KEV list, or is there a public proof-of-concept? Either one moves it to the front of the queue.
  5. What does the CVSS vector say? The score is a summary; the vector string shows whether it needs authentication, user interaction or network access.

For a worked example of this process on a real flaw, see our walkthrough of the PHP RCE vulnerabilities and how to check, patch and restart safely.

Why the exploitation window keeps shrinking

Defenders used to have weeks. Mandiant's time-to-exploit research, which tracks the gap between a vulnerability's disclosure and its first observed exploitation, shows how that changed:

  • 2018–2019: an average of 63 days.
  • 2021–2022: an average of 32 days.
  • 2023: an average of five days. Mandiant notes this figure excludes statistical outliers; without the exclusion it would be 47 days. See their 2023 analysis.

More recently, Google's M-Trends 2026 report puts the average time-to-exploit below zero: on average, exploitation starts before a patch exists. That is largely a zero-day story, and you cannot patch what has not been fixed yet, but it still changes what fast has to mean for the cases where a patch does exist.

Exploitation is also a leading way in. Verizon's 2025 Data Breach Investigations Report found vulnerability exploitation at about 20% of initial access vectors, up 34% year over year, with edge devices and VPNs patched slowly: a median of 32 days to remediate, and only about half fully fixed within the year.

Where AI changes the picture

Turning a vulnerability advisory into a working exploit used to take skilled people days. Several recent results suggest AI is compressing that:

Treat these with the right amount of salt. Proofs of concept in a lab are not the same as reliable attack campaigns, and most of this evidence is self-reported. But the practical conclusion does not depend on the exact numbers. A published patch is also a roadmap: attackers can diff the fix to find the bug, and tooling that automates that step lowers the skill and time needed. The gap between a fix is published and someone is probing for the unpatched version is getting shorter, so your patch cycle needs to be shorter too.

The same tooling helps defenders find and fix flaws faster. That is the other side of the same shift, and one more reason not to be the team still running last year's release.

Always run the latest version is the wrong rule

The instinct is right, but the rule needs one adjustment. On production systems you want to apply security patches promptly and plan major upgrades deliberately:

  • Patch releases within your supported branch (PHP 8.4.26 → 8.4.27, a point release of Node, a distro security update) are small, low-risk and should ship fast.
  • Major version jumps (a framework's next major release, a new database version) deserve staging and testing. Running whatever is newest on production is how you swap a security risk for an outage.
  • Stay on a supported branch. If your runtime or framework version no longer gets security fixes, every new CVE against it stays open forever.

A patching routine that holds up

1. Know what you run

You cannot assess a CVE against software you have not inventoried. Keep a list of runtimes, frameworks, system packages and the lockfiles for each project.

php -v
node -v
composer show --direct
npm ls --depth=0

2. Get advisories without relying on the news

  • Turn on GitHub's Dependabot alerts (or your Git host's equivalent) for every repository.
  • Run the audit tool for your ecosystem in CI and on a schedule:
composer audit
npm audit --omit=dev
pip-audit
bundle audit check --update
  • Subscribe to your OS vendor's security announcements, and to the advisory feed for any server software you run directly (web server, database, language runtime).

3. Triage with KEV first, CVSS second

A simple order that works for small teams:

  1. On the KEV list or with public exploit code, and exposed to the internet: patch within a day or two.
  2. Critical or high severity, and exposed: patch within a week.
  3. Everything else: batch into your normal release cycle.

Those windows are a starting point, not a standard. Adjust them to your risk and your team size, and write them down so the decision is made before the incident, not during it.

4. Automate the OS layer

For Debian and Ubuntu servers, security updates can apply automatically:

sudo apt install unattended-upgrades
sudo dpkg-reconfigure --priority=low unattended-upgrades

Our Linux server hardening checklist covers this alongside SSH hardening, firewall rules and user isolation, and the Fail2Ban guide adds brute-force protection on top.

Patching is a deployment problem

Most teams that patch slowly are not ignoring advisories. They are afraid of the deploy: manual steps, no easy way back, and a fear that just updating the dependency will break production on a Friday afternoon.

That is a deployment pipeline problem, and it is where a Git deployment automation setup earns its keep. The loop for a security patch becomes:

  1. Bump the dependency or runtime version on a branch and let your tests run.
  2. Merge, and let the push trigger the deploy.
  3. Rebuild with the updated lockfile through a build pipeline, so the patched packages are installed in a clean environment instead of on your live server.
  4. Ship with zero downtime deployments, so applying a patch does not mean a maintenance window.
  5. If the patch breaks something, use one-click rollback to return to the last good release while you investigate.

To be clear about the boundary: DeployHQ does not scan for vulnerabilities. Use the audit tools above for that. What it does is make shipping the fix routine and reversible, which is what makes teams willing to do it quickly.

Want patches to ship as a routine push instead of a risky event? Try DeployHQ free and set up automatic deployments with one-click rollback in an afternoon.

Patching is one layer, not the whole defence

Speed matters, but it is not sufficient on its own. A few habits limit the damage when something slips through:

FAQ

What does CVE stand for? Common Vulnerabilities and Exposures. It is a standard way to name and track publicly disclosed security flaws.

Is a CVE the same as a vulnerability? Not exactly. A CVE is the identifier and record for a vulnerability. Not every vulnerability has a CVE, and a CVE does not tell you how severe the flaw is.

What is the difference between CVE and CVSS? A CVE identifies the flaw. CVSS scores its severity from 0 to 10. A high CVSS score does not mean a flaw is being exploited; for that, check whether it is in the CISA KEV catalogue.

How quickly should I patch a critical CVE? It depends on exposure and evidence of exploitation. If the flaw is exploited in the wild and affects an internet-facing system, patch within a day or two. Otherwise a week is a common target for critical issues. Write your own policy down and follow it.

Should I always upgrade to the newest version? Apply security patches within your supported branch quickly. Test major upgrades in staging first, and make sure you stay on a branch that still receives security fixes.

Where can I look up a CVE? The CVE Program site and the NIST National Vulnerability Database list records, and your software vendor's advisory page usually has the clearest fix instructions.


Questions about shipping fixes faster? Email us at support@deployhq.com or find us on X at @deployhq.