In October 2026, attackers chained three zero-day vulnerabilities in Ivanti Connect Secure and used them to gain root privileges on the device without logging in. Connect Secure is software for remote access to corporate networks. The vendor says a limited number of customers were affected. Fixes for all three flaws are available.
The attack chain
The security firm Volexity describes three vulnerabilities, each enabling its own step. The entry point is CVE-2026-104935 (CVSS 8.6, EPSS score 88.94). Volexity attributes both the authentication bypass and the remote code execution to this flaw. A remote attacker can use it to run arbitrary commands without logging in. It has been considered actively exploited since 13 October.
To get to root level, the attackers used CVE-2026-111357 (CVSS 7.0, EPSS score 0.53, rated 8.8 by Ivanti itself). A critical function lacks authentication there, so a local attacker with a valid account can gain root privileges. On top of that comes a flaw that lets files be uploaded to a specific directory. This is most likely CVE-2026-118527 (CVSS 4.3, EPSS score 0.82), which Ivanti lists under the keyword predictable filenames. Volexity gives no identifier for this flaw, so the match is not explicitly confirmed. CISA added CVE-2026-111357 and CVE-2026-118527 to its KEV catalog on 30 October, which confirms active exploitation.
What Volexity observed
According to Volexity, all three flaws were zero-days. The researchers saw attacks from 3 October 2026 on a fully patched device running version 22.8R2.4. The attackers, which Volexity tracks as UTA0576, used a service account on the VPN device to access Active Directory credentials, local administrator passwords and BitLocker recovery keys. They also deleted entries from log files and left no web shell behind. Volexity does not say whether criminal or state-backed actors are behind it.
Ivanti says it discovered part of the chain itself and fixed it in July with version 25.1.3.0. The fixes for CVE-2026-111357 and CVE-2026-118527 only followed on 29 October, after the exploitation became known.
Affected and fixed versions
- CVE-2026-104935: affected are 25.1.1.1 and earlier, 25.1.2.2 and earlier, and 22.8R2.4 and earlier. Fixed in 25.1.1.2, 25.1.3.0, 25.1.3.1 and 22.8R2.5.
- CVE-2026-111357: affected are 25.1.1.2 and earlier and 22.8R2.5 and earlier. Fixed in 25.1.1.3 and 22.8R2.6.
- CVE-2026-118527: affected are all versions before 25.1.1.3 and before 22.8R2.6. Fixed in 25.1.1.3 and 22.8R2.6.
What to do
Ivanti recommends updating to version 25.1.3.0 or newer. It only runs on the new platforms ISA4500, ISA6500 and ISA8500. For hardware, that means you need new physical servers. For virtual or cloud appliances, you set up a new instance from the 25.1.3.0 template. If you keep running existing devices, you can update to 25.1.1.3 or, on the 22.8 line, to 22.8R2.6. Both versions close all three flaws. The 22.7 and 22.8 lines reach end of life on 31 January 2027.
Do not rely on Ivanti's Integrity Checker Tool. According to Volexity, it does not reliably detect these attacks, because the attackers only modified program code in memory and restored the configuration afterwards. A clean result therefore does not prove the device is intact. If you suspect a compromise, have network and memory forensics done. Volexity advises checking requests to /dana-na/auth/url_default/welcome.cgi whose X-DSCGI-PATH header does not match the requested path. It is also worth looking in the /tmp directory, where the attackers dropped files such as ml.py, rr.py or spray_hits.json. Since the actors harvested credentials, check and renew the affected accounts.