Trinetri logo
← Back to Blogs

How Long Is Your Exposure Window?

By Kishan BhanderiNovember 16, 20268 min
How Long Is Your Exposure Window?

Attackers now exploit serious vulnerabilities about a week before patches exist, while defenders take 43+ days to fix them, leaving you exposed far longer than your patch metrics show. Learn how to measure your true exposure window and five ways to shrink it.

How Long Is Your Exposure Window?

Ask most security teams how fast they patch and you will get a number. Thirty days for criticals. Fourteen if the vendor says it is being exploited. Maybe a dashboard showing mean time to remediate trending down quarter over quarter.

Ask a different question, though, and the room usually goes quiet: how long were you actually exposed?

Not how long it took to patch once you knew. How long attackers could have used the flaw against you, from the moment it was weaponized to the moment the last vulnerable system in your estate was fixed. That span is your exposure window. In 2026 it is longer than most organizations think, for a reason that is uncomfortable to admit: it now often starts before the patch exists.

This post looks at what the latest data says about both ends of that window, why the traditional patch cycle no longer fits the threat, and what practical teams are doing to shrink it.

Exposure window vs. time to patch

The exposure window is the time from when a vulnerability is first weaponized to when the last affected asset in your environment is remediated. It is not the same as time to patch, and the difference matters.

Time to patch usually starts the clock at disclosure, or when your scanner first flags the issue. It stops when most systems are fixed, and it is often reported as a median. That makes it a useful operational metric, but it measures your process, not your risk.

The exposure window starts the clock earlier and stops it later:

  • It starts at weaponization. If attackers had a working exploit a week before the advisory, that week counts, even though you could not have patched it.
  • It ends at the last fix. A median tells you when half your systems were patched. Attackers only need the one appliance that stayed vulnerable for eight months.

Qualys has formalized this idea as an Average Window of Exposure, measured across an environment from weaponization to remediation. Its example is Follina: weaponized 30 days before disclosure, with average remediation closing on day 55, for a window of about 85 days. Most of that window sat outside what a typical patch SLA measures: the stretch before disclosure and the long tail after the median.

The front edge: attackers moved earlier

Mandiant's M-Trends 2026 puts the mean time to exploit at negative seven days. On average, the most serious vulnerabilities are being exploited about a week before a patch is available. In 2018 the same measure was roughly 63 days; it crossed zero in 2024.

A negative average does not mean every flaw is exploited before disclosure. It means that among the vulnerabilities attackers care most about, which are exactly the ones your patch program prioritizes, pre-patch exploitation is now normal rather than rare.

Other datasets point the same way:

  • VulnCheck tracked 884 entries in CISA's Known Exploited Vulnerabilities (KEV) catalog with first-time exploitation evidence in 2025. Nearly 29% were exploited on or before their CVE was published, up from 23.6% the year before.
  • CrowdStrike's 2026 Global Threat Report says zero-day exploitation before public disclosure rose 42% year over year.
  • Once inside, attackers move fast. CrowdStrike puts average eCrime breakout time, from initial access to lateral movement, at 29 minutes. The fastest case took 27 seconds.

AI is a big part of why this keeps accelerating. Researchers showed back in 2024 that a general-purpose model could exploit most of a small set of one-day vulnerabilities for a few dollars each. Models have improved a great deal since then. Patch diffing, where attackers compare a fix against the old code to work out exactly what it closed, used to take skilled analysts days. It is now increasingly automated, so publishing a patch can itself start the race.

The back edge: defenders are not keeping pace

While attackers have moved earlier, remediation has not sped up to match. In some datasets it has slowed down.

MeasureFigureSource
Median time to patch, H1 202643 days (up from 32)Dataminr, citing Verizon and CrowdStrike data
Median time to patch known-exploited vulnerabilities87 daysMoody's, using Bitsight data
Median time to fix network appliancesOver one yearMicrosoft Security Research
Median time to fix Windows workstationsAbout two weeksMicrosoft Security Research
Average remediation, Cisco IOS XE flaw263 daysQualys

The Moody's figure is the most striking. Even for vulnerabilities known to be exploited, half are still open after three months.

The gap between endpoints and infrastructure is just as telling. Laptops and workstations sit in well-oiled patch pipelines and get fixed in about two weeks. Network appliances, VPN concentrators, firewalls and gateways are a different story. Patching them often means a maintenance window, a reboot of something business-critical, or a firmware upgrade nobody wants to own. Qualys found that for Cisco IOS XE even the median fix took 232 days, while endpoint medians stayed under 14.

That is a problem, because those edge devices face the internet and are exactly where attackers want to start. The systems with the longest exposure windows are often the ones with the most exposure.

Putting the two edges together

Put the two trends side by side and the shape of the problem is clear. As a rough illustration: if exploitation starts about seven days before disclosure and the median fix lands around day 43, a typical serious flaw leaves you exposed for about 50 days. That number combines figures from different studies, so treat it as a sketch rather than a measurement. The real cases are often worse.

  • Spring4Shell was exploited two days before disclosure. The average enterprise took 266 days to remediate it.
  • Cisco IOS XE was weaponized about a month early. The average fix closed at day 263.
  • Follina had an exposure window of about 85 days. Roughly 36% of it came before disclosure and 44% came from the long tail of slow patching. Together those two edges made up 80% of the window, and neither shows up in a typical median time to patch.

This is also why averages and medians tell different stories. Qualys found Spring4Shell's average remediation time was 5.4 times its median. The median describes your well-run majority. The average is dragged out by the forgotten appliance, the legacy server owned by another team, and the system nobody is allowed to reboot. Attackers do not care about your median. They scan for the stragglers.

Five ways to shrink your exposure window

You cannot patch something before a patch exists, and you will never patch everything on day one. Shrinking the window means working both edges: acting before the patch, and closing the tail after it.

1. Prioritize by exploitation, not just severity

A CVSS score tells you how bad a flaw could be. It does not tell you whether anyone is using it. Rank work by evidence of exploitation first: a CISA KEV listing, vendor reports of in-the-wild use, and threat intelligence relevant to your sector. Then use exploit-prediction scores such as EPSS to rank the rest. A medium-severity flaw being actively exploited on an internet-facing system should jump the queue ahead of a critical one buried on an isolated network.

To make this work, give exploited vulnerabilities their own fast lane. That means a separate SLA measured in hours or days, not the standard 30-day cycle.

2. Have compensating controls ready before you need them

With a negative time-to-exploit, some of your exposure will always fall before a fix exists. The answer is mitigation you can deploy fast:

  • Virtual patching through WAF or IPS rules
  • Network segmentation that limits who can reach a vulnerable service
  • Disabling the affected feature, module or endpoint
  • Restricting management interfaces to trusted networks only

The key is preparation. Build a playbook in advance that maps common vulnerability classes to the controls you can push within hours. Get those changes pre-approved. When the next advisory lands with "exploitation detected," you want a runbook, not a meeting.

3. Treat edge devices as tier zero

VPNs, firewalls, load balancers and gateways face the internet and have the slowest patch times in the data. Treat them accordingly:

  • Keep a complete, current inventory of every edge device and its firmware version.
  • Give them the tightest SLA in your program, not the loosest.
  • Plan maintenance windows in advance so a firmware update is routine, not a crisis.
  • Assume a vulnerable edge device may already be compromised, and hunt for signs of it after patching.

4. Measure the whole window, not just the median

What you measure is what you improve. If your only metric is median time to patch, your program will optimize the easy majority and ignore the tail. Add metrics that capture the full exposure:

  • Days exposed per vulnerability: from first known exploitation to last asset fixed.
  • Percentile fix times: report the 90th and 99th percentiles alongside the median.
  • Open exploited vulnerabilities by age: how many KEV-listed flaws are still open after 7, 30 and 90 days.

5. Automate what you can, and own the rest

A monthly patch cycle cannot keep up with exploits that arrive in hours. Automate deployment for the predictable majority, such as workstations, standard servers and common applications. That frees people to work on the hard cases. Then give every remaining asset a named owner, because the long tail is mostly systems where nobody is sure who is allowed to touch them.

How to measure your own window

You do not need a new platform to get a first answer. Pick the five most serious vulnerabilities your organization dealt with in the past year, ideally ones listed in CISA KEV, and work through each:

  1. Find the weaponization date. Use the earliest credible report of in-the-wild exploitation: the vendor advisory, the KEV entry, or threat intel reports. If exploitation came before disclosure, use that earlier date.
  2. Find your last-fix date. Look in your vulnerability management or patching records for the date the last affected asset was remediated or retired, not the date most were done.
  3. Subtract. The difference is your exposure window for that flaw.
  4. Split it into three parts: time before disclosure, time from disclosure to your median fix, and time from your median to your last fix. The biggest of the three tells you where to invest. Exposure before disclosure points to compensating controls. A slow median points to process and automation. A long tail points to ownership and edge devices.

Run this once and you will have a number you can track every quarter. More importantly, you will have a number you can put in front of leadership that describes risk the way attackers experience it.

The question that matters

The patch cycle was designed for a world where defenders had a head start. Disclosure came first, then a patch, then, weeks or months later, exploitation. That world is gone. For the vulnerabilities that matter most, attackers now often arrive first, and they find the systems we patch last.

That does not make patching less important. It changes what good looks like. Speed still matters, but so does being ready before the patch, closing the long tail after it, and measuring the whole span honestly.

So the question for your next security review is not "did we patch?" It is: how many days were we exposed, and what are we doing to make that number smaller?

How long is your exposure window? If you have measured it, I would like to hear what you found.

Stop the sprawl. Eliminate the guesswork.

Ready to See Autonomous Endpoint Management in Action?

Experience how Trinetri transforms endpoint chaos into clarity. Try the 30 minute demo session, our experts will walk you through how AI-powered automation, real-time visibility, and unified control can secure your entire hybrid environment, without the overhead.