Dev.to Security 🔐 Cybersecurity 👁 0 📖 3 min read

How fast do CVEs travel?

While writing yesterday's note, I realized that some malware sat on npm for six months before the advisory database most tools read caught up. So, I wondered... how fast does the news about a vulnerability travel? I pick

How fast do CVEs travel?

While writing yesterday's note, I realized that some malware sat on npm for six months before the advisory database most tools read caught up. So, I wondered... how fast does the news about a vulnerability travel? I picked 20 CVEs at random and split them up into two groups of 10: the "loud" ones (exploited in the wild: attackers are already using them) and the "quiet" ones (ordinary bugs in npm and PyPI packages).

This piece is a scrollable, animated chart on luisalima.com. Here are its cards, each with the chart at that step.

The chart, step by step

A CVE is an entry in a public catalog of known security bugs. Most of us start paying attention to a bug the day its CVE is published.

But let's take Acronis (CVE-2026-87886) as an example. The first post about it on X came out... 45 hours before its CVE record!

The first post on X comes from a big account: for 8 of 10 exploited bugs, the first post came from an account with more than 10k followers

News followed, 16 hours before the CVE record. And CISA (the US government's cybersecurity agency) had already listed it as exploited in the wild, the day before the record existed.

Only then did the CVE record land.

Is this unusual? No. Of 10 bugs exploited in the wild, 5 were public somewhere before their CVE record: on X, Reddit, in the news, on GitHub or on CISA's list.

Half the exploited bugs were public before their CVE (by channel)

Public exploit code came later, 15 hours to 22 days after the first mention on socials and news. But attackers didn't wait for public exploit code: for 4 of those 6, CISA had already listed the bug as exploited in the wild.

From first mention to public exploit: 15 hours to 22 days (median 11 days); for 4 of the 6, CISA had already listed the bug as exploited

Then I looked at 10 ordinary bugs in npm and PyPI packages. Before their CVE record, there was nothing. After it, there was a handful of posts and some aggregator news.

But... look left. Every one of them was fixed before its CVE: 25 days earlier for the typical bug, up to 73. So, the fix was public weeks before the record appeared. This is what we like;)

Fixed weeks before anyone said a word: 10 ordinary npm and PyPI bugs, days from the fixed version to the CVE record (median 25)

So the CVE record appeared after the attackers for the loud bugs, and weeks after the fix for the quiet ones.

What does this mean for you? For many bugs, the fix is out weeks before the CVE record that your tools are waiting for. So, how can we take fixes when they ship and not wait?

Weeks later, many installs are still vulnerable: share of last week's npm downloads on a vulnerable version, 5 packages

Being fast doesn't mean being careless (remember yesterday's note?). The update bot can wait a few days, CI can block install scripts, the scanners can check the lists (can be late, but always check) and the tests decide whether to merge. Even with a 7-day wait before updating, all 10 fixes would have landed before their CVE record.

Sometimes tests can't tell you if an update will break your app in unexpected ways, and the people who could check are overwhelmed. Use the agents! Agents can click through the change and record a walkthrough video, so a person approves in minutes instead of reading the whole diff. For backend, they can use curl and other tools and do the same thing.

So, have the systems in place so you can update fast, and remember that no single measure keeps your code and apps safe: it's defense in depth.

My methodology: 20 CVEs from 2026, picked before looking at any spread data. Loud: 10 bugs added to CISA's list of exploited bugs (KEV) between August and November, one per vendor. Quiet: 5 npm + 5 PyPI advisories reviewed by GitHub, not on KEV. For each, the first time it showed up on X, Reddit, Google News, GitHub, CISA's list, plus the fixed version and public exploit repos ("poc", "exploit" or "rce" in the name; I didn't open them, and fake ones that ship malware are common).

Caveats: n = 20: this is a field note. "First seen" is the earliest result a search API returned, so the real first post may be even earlier. CISA and many news dates are by day. The time axis stretches the first hours and squeezes the weeks.

Data via looot (disclosure: I'm a cofounder): X, Reddit, Google News, GitHub, OSV, CVE.org, NVD, CISA KEV.

Originally published on luisalima.com.

📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.