When every bug is a crisis, people stop listening

This morning, Andrew Thompson (@ImposeCost) wrote:

The collateral effects of excessive hype for otherwise not hype-worthy bugs. There’s serious attacks happening every day, and every minute that the most well positioned people to do anything about them are spent chasing “real but not crisis, but portrayed as crisis” work, the overall security situation worsens.

I have the same concern. We’re burning a lot of goodwill in a boy-cries-wolf pattern on bugs that are real, but not nearly as bad as the first claims or the nicknames suggest. I made a similar point in Groundhog Day?… don’t get beat up by the FUD in the news.

This week’s bugs still need to be patched. The packaging is the problem. Log4Shell-style nicknames and “critical” scores show up before the preconditions do.

Perhaps a better framing: real, but not a crisis.

On November 17, Calif disclosed CVE-2026-114938, named it Java4Shell, and scored it 9.2 Critical under CVSS 4.0. Oracle scored the same bug 8.1 (High) under CVSS 3.1, with high attack complexity. Calif says there is no universal exploit… an attacker has to find an application-specific way to get malicious Base64 into the vulnerable decoder, and the bug only hits the hardware-accelerated path on x86-64 CPUs with AVX-512 VBMI. ARM64 is not affected. Calif did demonstrate unauthenticated remote code execution against default installs of Jira, Confluence, Jenkins, and several other widely used applications, so if those face the internet on affected hardware, patch them now. Azul wrote, “Java4Shell is harder to weaponize quickly than Log4Shell, which gives defenders some time.” Azul also said the fix shipped in the scheduled November 17 update, with no exploitation in the wild reported that evening and the CVE not on CISA’s KEV catalog.

The next day, five NGINX memory-corruption bugs landed under the name NGINX2Root. F5 and nginx.org don’t use that name. F5 scored all five 9.2 Critical under CVSS 4.0. nginx.org rated two “major” and three “medium.” Successful code execution runs as the NGINX worker, not as root. Getting root requires chaining a separate, still-unpatched Linux kernel SCTP bug. The public Rapid7 proof of concept for CVE-2026-123007 is real, but it needs roughly 128 GB of HTTP/2 traffic and is unlikely to work behind an HTTP/2-terminating proxy. Censys reproduced it in a lab and got command execution as the worker, “not host root.” Wiz did confirm code execution, including an ASLR bypass, against official builds with HTTP/2 enabled, so internet-facing NGINX with HTTP/2 still deserves a prompt patch. As of November 19, Tenable had not seen reports of exploitation in the wild.

Log4Shell earned the panic. CISA lists it at CVSS 10, called it easy to exploit, and said it was being widely exploited. That’s the comparison these nicknames are borrowing, and most of them haven’t earned it. A recent case that actually was being exploited: Ivanti patched two more Connect Secure zero-days that had been chained in attacks since at least October 1, which I used as evidence in The boring list still wins (BleepingComputer).

When the next one shows up, I’d communicate it this way:

  1. Lead with exploitability and preconditions, not the nickname or the highest CVSS. Unauthenticated or not, internet-facing or not, and what has to be true for this to work.
  2. Say what you don’t know. Independently verified? Exploited in the wild? On KEV?
  3. If it is internet-facing and the preconditions match your environment, patch now. If not, schedule it and don’t page the executive team.
  4. Leave the marketing names out of internal notes and board briefings. Call it the CVE and the preconditions, not Java4Shell or NGINX2Root.
  5. Still patch, and match the urgency to the facts.

If you’ve had to walk a leadership team back from one of these or have a better way to frame and de-hype issues like these, I’d like to hear how you did it.

Leave a Reply