Latest

Cybersecurity

Ivanti patches two zero-days and leaves 5,000 devices exposed

Ivanti fixes two chained flaws present since October 1 in Connect Secure, allowing root without leaving malware and with 5,000 appliances exposed to the internet.

Ivanti patches two zero-days and leaves 5,000 devices exposed
Ivanti patches two zero-days and leaves 5,000 devices exposed

On Monday, November 2, Ivanti released patches for two new vulnerabilities in Connect Secure, its remote access gateway. They are CVE-2026-111357, a high-severity privilege escalation, and CVE-2026-118527, which allows files to be uploaded to an unpatched machine. None works on its own: for a month they have been chained with CVE-2026-104935, an authentication bypass fixed last month, to achieve code execution as root without credentials on devices that were fully patched.

The figure that dictates the urgency is given by Shadowserverwhich tracks more than 5,000 internet-exposed instances of Connect Secure and cannot say how many have already applied the updates. Connect Secure is not just any program on a company's network. It is the gateway through which employees and suppliers come in from outside, and that is why attackers have chosen it for years.

A device like this combines remote access, credentials and implicit trust in a single box that almost nobody checks daily. The uncomfortable detail of this chain is not the access, it is the absence of a trace: whoever got in left no malware or webshell, re-exploited the appliance with every command and tampered with the logs so that Ivanti's Integrity Checker Tool would see nothing unusual.

The chain that begins in October

Volexity traced an actor it calls UTA0576 and dated the use of CVE-2026-104935 to at least 3 October, combined with two other pieces that now have a name. The result was unauthenticated root access on updated devices, something that in practice turns the gateway into a silent entry point into the entire corporate network.

The intelligence firm Defused recorded the first signs of exploitation on 1 October, two days before the date Volexity published. That two-day difference does not change the incident, but it does serve as a reminder that attack timelines are built with a delay and that the first signals are usually in the traffic, not on the disk.

Ivanti's timeline explains why today's two vulnerabilities arrive late. When the company discovered the exploitation of CVE-2026-104935 it already had a fix for the privilege escalation in quality control, and it decided not to delay delivery of what cut off the main attack path.

It is a defensible decision and at the same time a lesson in how partial patches are communicated. Anyone who read the October advisory and updated was left with half the problem solved, without knowing it, for four weeks.

An attack without malware that erases its traces

The technical description of the intruder is the part that should most worry a defensive team. Instead of installing a backdoor, the attacker repeated the exploitation every time it needed to execute something. There was no binary to detect, no persistent process to kill, no file to send to a sandbox.

There were only seemingly legitimate requests against a service that was doing its job. Add to that the manipulation of records to evade the Integrity Checker Tool, the mechanism Ivanti uses to check whether a device has been tampered with. The tool is still useful, but it stops being proof of innocence when the adversary knows exactly what it looks at.

The lesson repeats itself in every incident of this kind. The integrity of a device is not proven by its own verifier alone; it is proven by external telemetry that no one inside the team can rewrite. In practice, that means keeping the gateway logs for weeks in a separate system and reviewing administrative sessions outside business hours.

It also means auditing certificates issued without a clear process behind them and treating any new access to the administration console as an event that deserves an explanation. It is less glamorous work than searching for a malware name on VirusTotal and, precisely for that reason, it is the work that tends to be missing when the post-incident audit arrives.

Five thousand exposed doors and a patch calendar

The number of exposed instances does not fall as fast as the advisories are published. Many organizations inherited their Connect Secure from an earlier contract, have it managed by an external provider or run it in a branch office nobody visits. Add to that the fact that the device works, and updating it lands on a to-do list competing with a hundred more urgent things.

On October 28, the US cybersecurity agency added CVE-2026-104935 to its catalog of exploited flaws and gave federal agencies until Friday to apply it. That kind of order has a useful side effect: it turns a vendor note into a deadline with consequences. Outside the public sector, most companies still lack an equivalent calendar.

Ivanti has a loaded track record in this family. In January 2024 it fixed a command injection, CVE-2024-21887, together with an authentication bypass, CVE-2023-46805, and two weeks later a server-side request forgery flaw. In January 2025 came CVE-2025-0282, a remote code execution exploited as a zero-day.

The pattern is no coincidence. These are devices with a large attack surface, little observability and a position in the network that makes them profitable. Anyone with one in front of them should treat the update as an incident rather than maintenance, with credential rotation and a review of access over the past six weeks.

The clock matters as much as the flaw. A patch that arrives four weeks late does not prevent the intrusion, it only shortens its duration, and in that interval the attacker has already decided what to copy, how to get out and which vendor to blame.

Applying the patch is the first step, not the last. The guide to reading a CVE and knowing whether it affects you helps to organize that review without wasting time on acronyms, and the rundown of the CVE-2026-98432 in the Linux kernel that we have already published shows just how far a single month can accumulate critical flaws across very different layers.

Why VPNs will remain the easy target

Attackers do not choose these gateways for technical elegance. They choose them because they concentrate three things no company can afford to lose at once: access, identity and trust. A flaw in an internal application affects one application. A flaw in the front door affects everything behind it.

That calculation will keep working as long as VPNs remain the glue of corporate networks. The real fix is not a faster patch, it is reducing the number of doors. Every service moved behind an access model with verified identity and least privilege takes value away from the next zero-day chain, which will come.

Meanwhile, the only realistic defense is boring and familiar: an inventory of what is exposed, patching on a schedule, logs out of the attacker's reach and an honest review of how much trust each device gives away. None of those tasks fits in a headline, and all of them together decide whether a zero-day chain remains a scare or becomes a lost quarter.

If the attacker can get in, execute as root and erase their tracks without leaving a single file, how many intrusions from recent weeks are still going undetected on networks considered clean because the antivirus said nothing?

José Maldonado

Author

José Maldonado

Activist and blogger on technology, free software and blockchain.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading