On Wednesday 18 November, at 19:17 UTC, CISA published version 2026.11.18 of its known-exploited-vulnerabilities catalogue. One of the four additions is CVE-2024-10441, a flaw in the operating system of Synology NAS devices whose patch shipped on 5 November 2024. Between those two dates there are 743 days: the fix had been downloadable for two years on the day somebody confirmed it is being used.
The textbook reaction is "update the NAS", and it is the right one: it is cheap and it needs doing. What the advisory does not answer is the question that actually orders your afternoon: where can that device be reached from? On a NAS, that answer is not in your firewall.
Three clocks for the same flaw
Every date in this case is public and easy to cross-check. Everyone has counted the 743 days; the stretch in the middle, which is the interesting one, we have not seen covered anywhere:
Those 134 days in the middle are the interesting ones. For four and a half months a vendor fix existed and there was no NVD record — and NVD is the list almost every vulnerability scanner drinks from. If your patching process starts when the scanner turns a light red, during that stretch your process depended on a file that had not been written yet.
We wrote about the other end of this rope when we took apart one ransomware family's entry points and found that not one of the five was from that year. This case is the same phenomenon from the defender's side: years can fit between the day a flaw becomes dangerous and the day it becomes news.
Two advisories, two different vectors
The same CVE has two public records, and they disagree on the one thing that actually orders the work: where it can be exploited from. Both scores have been sitting in the aggregators for a year and a half; what we have not seen anyone do is read the difference operationally.
A 9.8 and an 8.8 are, for prioritisation purposes, the same traffic light: both are red and both join the same queue. The difference between AV:N and AV:A is something else, because it changes the plan. If the attack is network-based, anyone who can reach the box will try it. If it is adjacent-network, the attacker has to be inside your segment already, and then this becomes the second step of an incident that started somewhere else. What you need to look at changes completely depending on which of the two records describes your box.
There is a boring and likely explanation, and it is worth stating because it stops you building a mystery where there is none: the ZDI advisory declares one single affected product, the DiskStation DS1823xs+, and it is credited to Ryan Emmons of Rapid7, who also appears in the acknowledgements of Synology's advisory. In other words, we are most likely comparing "what I tested on one specific box" with "what the vendor scores for the whole range". That turns the discrepancy into a concrete question about your model.
And there is a third data point, the heaviest one, which is in none of this week's coverage: in that same NVD record, CISA left its own SSVC assessment on 19 November at 04:55 UTC, with three fields — exploitation: active, automatable: yes and technicalImpact: total. That automatable: yes describes a flaw you can fire at many targets at once, which is the opposite of what requiring adjacency would suggest. The people applying the pressure are reading it the wide way.
One detail in the table that throws people if it is not explained: ZDI dates the vendor notification to 2 December 2024, almost a month after the patch. That is not an error, it is how Pwn2Own works — bugs are handed over at the contest and ZDI's advisories come out on their own schedule — but it is worth saying before someone reads it as an inconsistency.
Two descriptions of the same flaw, and they do not match
Advisory SA-24:20 covers three CVEs at once and has a single affected-products table, shared by all three. We already published this table on 1 November, when we went through the three incompatible accounts of an attack on Synology NAS devices and the problem of a NAS having two patch inventories. We are repeating it because today it is needed for what comes next:
Further down, the advisory does break out which versions each CVE touches, and that is where things stop adding up. The only critical flaw of the three is ours, and its description names BeeStation OS before 1.1-65374 and DSM before 7.2-64570-4, 7.2.1-69057-6 and 7.2.2-72806-1: three branches, no 7.1 and no 6.2. The other two do name 7.1, and one of them also 6.2; both are classified as moderate. Yet in the table above, the DSM 7.1 row is marked critical. Two parts of the same document, read on the same afternoon, do not give the same answer.
That breakdown was also not there from the start: the advisory's own revision history dates the "Disclosed vulnerability details" entry to March 2025. For the first four-odd months, anyone running DSM 7.1 was looking at a table with their branch flagged red and no way to tell which of the three flaws applied to them.
And there is a third document. The NVD record for this CVE holds two descriptions in the same API response. The English one names 'BeeStation OS (BSM)' and three DSM branches. The Spanish one names five DSM branches — 6.2.4-25556-8, 7.1.1-42962-7, 7.2-64570-4, 7.2.1-69057-6 and 7.2.2-72806-1 — plus Synology Unified Controller (DSMUC) before 3.1.4-23079, a product the English version never mentions; and it calls the BeeStation one 'BeeStation Manager' along the way. That is two different scopes for the same identifier, in the same record, verifiable with one API call.
In practice it resolves the same way: you upgrade to the fixed release for your branch and you are done. What does not resolve itself is the other half of the job — knowing what you have and where it is plugged in — because no advisory does that part for you.
The device that reaches the internet without you opening a port
When a company inventories its exposed surface, the usual move is to look at inbound firewall rules and router publications. No rule, no exposure. With a NAS that arithmetic fails, and it fails because of how the product is designed.
Synology's own documentation spells it out: QuickConnect lets you connect to the NAS over the internet "without the hassle of setting up port forwarding rules". The product white paper describes three attempts in sequence: a direct connection over LAN or WAN, if the NAS is reachable by IP or domain; failing that, hole punching assisted by Synology's servers, which exploits the fact that home routers do not block outbound traffic; and, as a last resort, the relay service, where "the QuickConnect Server will inform the NAS to create a virtual tunnel between the NAS and the Relay Server" and a port is assigned on that relay whose traffic is redirected to the device.
It is worth not stretching this further than it goes. QuickConnect does not make anyone adjacent: the AV:A in the vector describes someone sharing a network with the device, and a relay publishes a service, not a segment. What QuickConnect does is something else, and it is enough: it gives the NAS an inbound path from the internet that your perimeter inventory does not record, because there is no inbound rule to record. If you build your list of exposed things by reading inbound rules, that device will not be on the list even while it is answering half the world.
The fifteen-minute check
All of this can be checked without installing anything and without stopping the service. In this order:
- The version, hyphen included. "I'm on 7.2" is not enough. Write down the full string and compare it with your branch's row. A
7.2.2-72806without the trailing-1is not the fixed release. - External access. In the control panel, if QuickConnect is on, record it as exposed surface even though your firewall has no inbound rule at all. Same for any DDNS you have set up.
- The router, separately. Port forwards and, above all, the UPnP table: that is where you find publications no human decided on.
- Who is an administrator. Privileged local accounts, how many there are, and when each last logged in.
- Where the logs go and how long they last. This is the one that stings, and we explain it below.
- What hangs off the NAS. If the same box holds the shared folders and is the backup target, you have a single point whose failure is an outage, not a fault.
If they ask for evidence, somebody has to be able to produce it
CISA's record carries the forensic-triage field set to Yes: patching is not enough, you have to look first at whether they are already in. It is a reasonable demand and, on a closed appliance, a fairly awkward one to meet. We ran into this with two kernel flaws whose patch was a reboot — precisely the operation that erases part of what you were asked to collect.
The useful decision there is not made today; it was made the day you set the box up. If its logs ship over syslog to somewhere the NAS itself does not control, you can answer. If they live only inside, with short rotation, on the same volume whoever got in could have touched, the honest answer to "were you compromised two years ago?" is "there is no way to tell", and it is better said on day one than on the last day.
When this is not urgent for you
If your NAS sits on a VLAN with no internet egress, no QuickConnect, no DDNS and no router publications, and the only path to it runs through the office network, this is one more line in this month's maintenance window. Upgrade when it suits you and get on with your life. No crisis meeting. And if your backups also leave the box for somewhere the box cannot erase, you did the hard work long ago.
The position that does not work is the middle one, which is also the most common: a box that has been powered on for years because "it works fine", with remote access switched on one day so somebody could work from home, administered by someone who no longer works here, and that nobody reboots because everyone's work lives inside it. That NAS does not have a patching problem. It has an ownership problem.
What we are not claiming
We do not know who is exploiting this or against whom: CISA publishes no victims and the ransomware-campaign field reads Unknown. We have not reproduced the flaw and we have no proof of concept of our own. We do not know how many devices are still unpatched, because any figure floating around comes from mass scans that cannot tell a business NAS from a home one. And we are not a Synology partner or reseller: we operate whatever our customers have, whoever made it.
What we will stand behind, and what we take away from today: the catalogue has not discovered a new flaw, it has changed the status of one that has been fixed for two years. The only thing that decides whether this news affects you is a question no list will answer for you: who can open a session against that device today, and by which path? If your answer is "nobody, it has no open ports", look again. It may not need any.
Sources (verified at origin): the date added, the 21/11/2026 deadline, the forensicTriage: Yes field, the knownRansomwareCampaignUse: Unknown value, the short description and the CWE-116 classification come from the JSON feed of CISA's Known Exploited Vulnerabilities catalogue, version 2026.11.18 released at 19:17:41 UTC with 1,786 entries, downloaded and read by us. The publication date (05/11/2024), the last-updated date, the severity and the fixed-release table — advisory Synology-SA-24:20. The NVD publication date (19/03/2025), the description scoping the affected branches, and the 9.8 score with AV:N vector attributed to Synology as CNA — NVD record, queried via API. The 8.8 score with AV:A vector, the technical description of the username used to build environment variables, execution as root, and the notification and publication dates — advisory ZDI-25-214. The sentence about port forwarding — DSM help; the three sequential attempts and the virtual-tunnel quote about the relay server — QuickConnect white paper. CISA's SSVC assessment of 19/11/2026 at 04:55 UTC (exploitation: active, automatable: yes, technicalImpact: total) sits in the same NVD API response. Translations of the quotes are ours. The 743 days have been counted by all the coverage; ours are the arithmetic of the 134 days, the operational reading of what the vector changes, the inconsistency between the critical DSM 7.1 row and the per-CVE breakdown, the contrast between NVD's two descriptions, and the fifteen-minute check. Photograph by Shixart1985, CC BY 2.0 via Wikimedia Commons.
Do you know which of your devices reach the internet without asking?
Inventorying what is genuinely reachable, not just what has a firewall rule, is half an hour of work that saves you two-year-old surprises. We do it as part of cybersecurity, and the principle of not trusting the network a device sits on is Zero Trust. If you want us to look at yours, write to us.
Talk to everyWAN