Between October 26 and November 4, the United States Cybersecurity Agency added twelve vulnerabilities from ten vendors to its CISA catalog of exploited vulnerabilities. Seven arrived with a three-day deadline for federal agencies, four had a score of 9 or higher and three still had no linked fix in the sources that track them. In October, the same agency had added 33 in a full month.
The change is not in the number, it is in the haste. A three-day deadline forces deployment with no room to test, and that favors products that update themselves and penalizes those that need a maintenance window. For a federal agency, the difference between three days and fourteen is not one of calendar, it is one of process: without time to test in a replica environment, the patch is applied blindly or not at all.
The story that best explains what that pace means is not the list, but what happened at a specific GitLab installation. Its maintainers detected two attempted exploits, saw eight workflows interrupted and rebuilt the entire machine without having confirmed the damage.
The catalog numbers
The catalog of known exploited vulnerabilities exists for one reason: to separate the flaws someone is actually using from those that are only serious in theory. A vulnerability is added when there is evidence of real exploitation, and that evidence can be a vendor advisory, a report from researchers or a documented incident.
Of the twelve additions in the period, four arrived with a critical score and three had no patch located by those tracking them, according to the tally published by CyfencedLists of that kind are the ones that generate the most pressure inside an organization, because the alternative to the patch is almost always shutting down a service.
The short deadline is not new either. What changed this autumn is how often the agency applies it, and that turns an exception into the routine of an operations team.
GitLab, the 10.0 almost nobody was watching
The CVE-2026-97229 vulnerability affects GitLab Community and Enterprise from 19.2 to 19.4, and was fixed in 19.4.2. The cause is an integer overflow triggered by a specially crafted regular expression in the continuous integration configuration, and it allows an unauthenticated attacker to execute code on the server.
The score is 10 because the vector is network, the complexity is low and no user interaction is required. What is unusual about this case is that the flaw reached the catalog because of an incident reported by the Alpine Linux project, not because of a broad campaign.
GitLab is the place where a company's code lives, along with the credentials that deploy it. An attacker who executes code there needs nothing more to read repositories, modify artifacts or sign a version that no one has reviewed.
The attempts against the Alpine instance occurred on October 20 and 21, with specially crafted continuous integration configurations. Workflows were anomalously interrupted eight times and, in four of those cases, the exploit did not work in the next phase. The repeated attempts on October 25 and 26 were blocked, and although they found no evidence of a successful compromise, the maintainers considered the instance compromised and rebuilt it from scratch, a criterion that Help Net Security has documented in similar incidents from the same week.
Cisco and the restart without credentials
The other notable addition in the period is CVE-2026-76418, a denial of service with a score of 8.6 affecting Cisco's ASA and FTD firewall software. The flaw lies in request processing when secondary authentication is configured in a tunnel group of the SSL VPN service.
An unauthenticated remote attacker can cause an unexpected restart of the device. The detail that widens the attack surface is that there is no need to complete a VPN login to reach the vulnerable path, so anyone who can send traffic to that interface can take the device down, according to the review published by CyberSecureFox.
On a perimeter network, a repeated restart amounts to a service interruption and, in some configurations, to a window in which the rules are not applied as they should be. That turns an availability flaw into a confidentiality problem when the restart coincides with a maintenance task that no one scheduled.
Three cases without a patch
The three cases without a patch are the most uncomfortable. The best known is PHP2Shell, two vulnerabilities that allow code execution before the application starts, fixed in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 on October 21. The problem is that PHP 8.1 and earlier no longer receive support, and the agency went as far as recommending that servers still on those versions be taken off the internet, a decision many organizations cannot make in three days because the system supports a production service.
That is where the limit of any catalog shows. A list of mandatory patches works when the patch exists and fails when the only way out is to shut something down, so mitigation ends up being a business decision rather than a security one.
The Brotli case follows the same logic. The fix is in library 1.2.1 and in Node.js versions 22.23.4, 24.22.1 and 26.12.1, as well as .NET 8.0.33, 9.0.22 and 10.0.14, while Node.js 20 is left without a patch. The Brotli version that Node reports does not change after updating, so checking it does not tell you whether you are protected.
What to do with the list on Monday
The list is not a roadmap, but it is a source of priorities. A small team can use it in three concrete ways without spending more than an hour a week on it. First, cross-check the twelve names against the inventory and see which ones exist in production. Second, check whether the three without a patch are exposed, because those are the ones that require immediate mitigation. Third, review whether any product updates itself, because in that case the three-day deadline is met or missed without anyone intervening.
The underlying consequence is that the window between the publication of a patch and its exploitation has narrowed to the point of disappearing in some cases. When a flaw enters the catalog it is because it is already being used, so the time to decide is counted in hours.
That pace explains why the kernel CVE triage has become a job in itself. It also explains why cases involving network devices such as the two Ivanti zero-days are now resolved with emergency patches instead of scheduled windows.
For an organization without a dedicated team, the only sustainable response is to reduce the exposed inventory and automate what can be automated. The list of twelve in ten days is not a peak, it is the new size of the weekly task, and the example of the software that powers 70% of the web with a single employee remember that the other half of the problem is not up to you.