Read three outlets on the Citrix NetScaler attacks next to the vendor research behind them and you get at least four start dates. The earliest, Aug. 21, comes more than five weeks before the Sept. 27 advisory that most defenders used to start their clocks. The gap between those dates is the finding, and it changes where a compromise assessment has to begin.
What the three accounts say
Citrix disclosed eight vulnerabilities in NetScaler ADC and NetScaler Gateway on a Sunday, two of them confirmed as exploited, according to the UK’s National Cyber Security Centre. The two are CVE-2026-88771, which allows an unauthenticated remote attacker to execute arbitrary commands, and CVE-2026-88772, a memory buffer flaw that leads to remote code execution or denial of service. Three outlets then covered what followed, and each told a different part of it.
CyberScoop: the gap before anyone knew
CyberScoop built its report around detection time. It reported that Mandiant dates the earliest known exploitation of CVE-2026-88772 to Sept. 3, and that attacks were not confirmed until late the week before the advisory. It added that GreyNoise has seen CVE-2026-88771 exploited since at least Sept. 24, with the campaign likely starting earlier. Mandiant’s Charles Carmakal counted “dozens” of affected organizations, per the same article, and Mandiant told CyberScoop its timeline could still move as responders find more evidence.
SDxCentral: the disruption
SDxCentral led with fallout. Citing Techzine, it reported that the Netherlands’ Ministry of the Interior took all its Citrix environments offline over the weekend, that patients of two major Dutch hospitals were unable to view their records, and that a third hospital, Frisius MC in Leeuwarden, shut down some of its digital systems as a precaution. It noted that only Dutch organizations had publicly reported issues caused by the flaws so far. The timeline reaches back further than any of the three accounts: Palo Alto Networks’ Unit 42 threat brief says the earliest activity it identified occurred on Aug. 21, when two hosts sent requests consistent with version fingerprinting to a NetScaler Gateway appliance at a US organization, then sent the same requests to more than 100 other systems over Aug. 21 and 22.
The Hacker News: a third flaw arrives
The Hacker News covered the part the other two could not, because it happened later. On Oct. 5 it reported a patch for CVE-2026-88779, a memory overflow with a CVSS score of 8.7 that can cause denial of service when NetScaler is configured as a SAML service provider or identity provider. Citrix said it had observed targeted attacks on unmitigated deployments and had not identified an impact on the integrity of customer data. The outlet reported that CISA added the flaw to its Known Exploited Vulnerabilities catalog with an Oct. 7 federal deadline.
Where the accounts differ
The start dates differ, and they are not all measuring the same thing. Unit 42’s Aug. 21 is version fingerprinting. Mandiant’s Sept. 3 is the earliest known exploitation of one specific CVE. GreyNoise’s Sept. 24 is the earliest sighting for the other. Each is a floor set by what that vendor’s telemetry could see, and none of them is a confirmed first day.
The framing differs too. SDxCentral places the damage so far in Dutch public institutions and hospitals. CyberScoop quotes Carmakal attributing the attacks to “advanced and suspected state-sponsored threat actors” and expecting “broad and opportunistic exploitation” of both zero-days by a variety of threat actors in the near term. Those readings call for different postures: the first says check whether you were selected, the second says assume you will be scanned.
On the basics the three agree. All describe flaws exploited in the wild, and all report fixes that arrived after attacks had begun.
The sequence of warnings matters as well. CyberScoop reports that the first unconfirmed rumors of the attacks surfaced nearly two days before Citrix disclosed the flaws on Sunday. Defenders therefore got their first signal on different days, and the date a team treats as day zero often depends on which of those signals reached it.
What the coverage adds up to
No single article says what the three say together: the advisory date is the worst place to start a hunt. A team that began its review on Sept. 27 would have looked at the days after the first confirmed exploitation and missed at least five weeks before it.
The reason sits in the appliance class. CyberScoop quotes Mandiant on why attackers favor these devices: “Because most edge devices do not support endpoint detection and response (EDR) monitoring, targeting them, particularly through exploiting zero-day vulnerabilities, provides threat actors with an infection vector that is difficult to detect and prevent, and the opportunity to scale a campaign as long as the exploit remains undiscovered.” CyberScoop adds that edge devices accounted for 48% of enterprise zero-days last year, per Google Threat Intelligence Group. With no endpoint agent on the box, logs from before the advisory often do not exist, so the only evidence of an early intrusion may be what the attacker left behind.
That is why the dates in the reports are minimums. Each vendor is reporting the earliest thing it can prove, and the proof depends on who had sensors in front of which customers. Our read is that a lookback anchored to the earliest published date, with extra margin, is the defensible choice until responders stop revising their timelines.
Why the exposure is not evenly spread
The sectors named in the coverage depend on this appliance in different ways. In the Netherlands, patients of two major hospitals lost access to their records and Frisius MC shut down some digital systems as a precaution, according to SDxCentral, which shows what taking these systems offline can cost. Our read is that the sectors Mandiant listed (government, financial services, education, telecom, legal and professional services) should check their own log retention against the Aug. 21 date instead of assuming it matches a peer’s.
What it means for the security leader
Patching closes the door and does not clear the house.
The NCSC advice says as much. It tells defenders to isolate affected systems and replace them with new, fully up-to-date systems where possible, investigate for compromise using the published indicators of compromise, and only then re-introduce the appliance. We made the case for that sequence in Replace an Exploited Edge Appliance Before You Trust It Again. The window question is the same one we raised in Attackers Probed Mail Servers Between a Zimbra Fix and Disclosure: attackers were on those servers before the fix and the advisory.
Three checks are worth finishing this week:
- Inventory NetScaler appliances running 14.1 before 14.1-73.37 or 13.1 before 13.1-64.23, the affected ranges the NCSC lists.
- Find the earliest date for which you hold authentication, VPN and web server logs from those appliances, and note whether they reach back to Aug. 21.
- Identify any appliance configured as a SAML service provider or identity provider. CVE-2026-88779 applies to those, and Citrix’s fix is 14.1-73.41 or 13.1-64.28 and later releases.
The defender close
Treat Aug. 21 as the working start date until a vendor publishes an earlier one. Pull every log you hold for NetScaler from that date forward, and investigate every server and system the appliance connected to. Change the service account passwords, secrets and user credentials that passed through the appliance and revoke the certificates stored on it, steps Citrix lists for suspected compromises, as The Stack relays them. If your retention window is shorter than five weeks, write that down as a risk, because it caps how sure you can be.
Source: NCSC alert on exploitation of Citrix NetScaler ADC and Gateway vulnerabilities