Citrix patches two NetScaler zero-days exploited in the wild for unauthenticated RCE

· CyberSecurity

TLDR

  • Citrix shipped its third emergency NetScaler fix in two weeks, for memory-overflow flaw CVE-2026-107406 (CVSS 9.5).13
  • Exploitation of recent NetScaler flaws followed public proof-of-concept code within hours, so patching should not wait.14
  • It does not yet appear to be exploited in the wild.13

Update, 9 October, what changed: Citrix shipped its third emergency NetScaler fix in two weeks, patching a newly disclosed memory-overflow flaw, CVE-2026-107406, with a critical CVSS v4.0 score of 9.5.13 The bug affects NetScaler ADC and Gateway appliances configured as a SAML identity provider or service provider, and "may lead to remote code execution or denial of service under specific configuration conditions," according to Citrix parent Cloud Software Group.13 Secure Private Access hybrid deployments running NetScaler instances on-premises with a Citrix Cloud control plane are also affected and must be upgraded.1314

Unlike the SAML-authentication denial-of-service bug patched on 4 October (CVE-2026-88779), which Citrix confirmed was being actively exploited, CVE-2026-107406 does not yet appear to be exploited in the wild.13 NHS England's National Cyber Security Operations Centre nonetheless assesses it is likely to be exploited, citing a pattern of mass exploitation "within hours" of proof-of-concept code going public for recent NetScaler flaws.14 Fixed versions are 14.1-73.46 and 13.1-64.29 and later, plus the FIPS and NDcPP equivalents; NetScaler 12.1 and 13.0 are end of life, do not receive the patch, and are likely still vulnerable.14 Citrix has already applied the fix across its own Citrix-managed cloud services and Adaptive Authentication.13

Update, 8 October, what changed: eSentire's Threat Response Unit identified four distinct threat-actor clusters exploiting CVE-2026-88771, some active before Citrix's 27 September disclosure, deploying a mix of custom and legitimate tooling across eight affected eSentire customers.11 Cluster A ran a pre-disclosure campaign as early as the beginning of September, hiding a PHP web shell inside a fake ".deb" package among legitimate NetScaler files, modifying Apache's configuration to execute it, and setting the SUID bit on /bin/sh for root-level access; eSentire said this web shell is nearly identical to the "lightweight installer web shell" family Google Threat Intelligence Group previously tied to the related CVE-2026-88772.11

Cluster B hit three customers roughly 24 hours after public disclosure, on 28 September, but rather than deploying malware it installed Platypus, a legitimate open-source fleet-management tool, giving attackers a shell, file management and network tunnelling; eSentire has not seen Platypus in prior incident data but noted it has appeared in other research on China-nexus actors, including Black Lotus Labs' work on an IoT botnet, and stressed any link to this cluster remains suspected, not confirmed.1112 Cluster C reached four customers on 29 September using a Perl script that added a hidden superuser account to NetScaler, deleted other accounts (possibly to lock out rival attackers), archived and exfiltrated the appliance's configuration folder, and planted a password-protected web shell disguised as a CSS file so requests to it blend into normal Apache logs; eSentire also spotted "LLM-style comments" in the script, suggesting the attackers used AI to help write it.12 Cluster D hit one customer the same day with a Python script that writes a reverse shell disguised as an SNMP service.12

Two of the four clusters used a public proof-of-concept, released by watchTowr on 28 September, within 36 hours of its release, with Cluster C even copying watchTowr's fake crash message; eSentire said the speed "shows how quickly threat actors target internet-facing appliances."11 eSentire has not formally attributed any cluster to a specific group.11

Update, 4 October, what changed: the unpatched bypass flagged on 3 October now has a CVE number and a fix. Citrix issued emergency updates early Sunday for CVE-2026-88779, a SAML-authentication denial-of-service bug with a CVSS score of 8.7, after attackers used it against unmitigated appliances to force repeated reboots.9 The flaw hits NetScaler ADC and Gateway appliances running SAML authentication with Gateway or AAA functionality; Citrix says it affects service availability only and has found no impact on customer data integrity, though researchers are still checking whether it can also be used for remote code execution.9 Anyone who already patched CVE-2026-88771 through CVE-2026-88778 must upgrade again, to 14.1-73.41 or 13.1-64.28 (or the FIPS/NDcPP equivalents), since those earlier builds remain vulnerable to this flaw.9

Citrix frames the crashing as "configuration-dependent" rather than a failure of last week's patches, saying the nsaaad authentication daemon crashes on specific, crafted SAML traffic and triggers the appliance's watchdog to force a restart; administrators can check exposure by looking for add authentication samlAction or add authentication samlIdPProfile in their configuration.10 watchTowr Labs, which first reproduced the earlier bypass, said it has also reproduced this vulnerability and has mitigation rulesets available to its Active Defense clients.10 Citrix is also distributing Global Deny Lists to block known malicious IPs, but says customers should still install the new patch as soon as possible.9

Update, 3 October, what changed: a fresh, unpatched exploit for NetScaler devices has been circulating since the evening of 2 October, apparently evading the fix for CVE-2026-88771/CVE-2026-88772 that shipped a week earlier.7 Security researcher Kevin Beaumont said one of his honeypots, already fully patched, began running a downloaded malware binary, concluding "both were patched, so new vuln."7 watchTowr Labs said it had "successfully reproduced this vulnerability," and the exploit is believed to work by sending a high volume of SAML requests to a vulnerable appliance.7 Citrix acknowledged the issue in a blog post but has not named an effective mitigation or shipped a patch, telling affected customers to contact support while it investigates.7

Separately, researchers reiterated that even fully patched appliances can remain compromised: Unit 42 warned that the September fix "will not remove access for attackers that have already established persistence," since the update does not clean up the altered Apache configuration, dropped web-shell files or changed permissions left behind by earlier exploitation of CVE-2026-88771 and CVE-2026-88772.8 Published telemetry now places the earliest reconnaissance in that campaign at 21-22 August and confirmed exploitation from 3 September, roughly three weeks before Citrix's 27 September bulletin.8

Update, 1 October, what changed: Security firm LevelBlue's Threat Hunt Operations & Research team said it identified malicious NetScaler authentication events across multiple customer environments containing attacker-controlled usernames weaponising CVE-2026-88771, consistently featuring variations of the strings "pitboss" and "NSPPE."6 Beyond basic vulnerability checks, LevelBlue observed attackers using curl or wget to fetch second-stage payloads from external servers, including a Python script that establishes a reverse shell to a remote IP over port 443 and kills processes tied to "/var/python/bin/customsnmpd," and a Perl script, update_c08937.pl, that creates a local superuser account named "sec_monitor" by modifying /flash/nsconfig/ns.conf.6

That Perl payload also archives the /flash/nsconfig directory, uploads the resulting configuration-data archive to an external server, then deletes the archive and erases itself to reduce its forensic footprint, before changing permissions on /bin/sh and deploying a PHP web shell at /var/netscaler/logon/LogonPoint/.local_journal.6 It modifies /etc/httpd.conf to enable PHP execution and maps the web shell to URLs resembling legitimate NetScaler CSS resources, corroborating activity GreyNoise had separately observed.6 LevelBlue said the techniques "demonstrate activity extending beyond basic vulnerability validation," spanning payload retrieval and execution plus collection and staging of NetScaler configuration data.6

Update, 30 September, what changed: GreyNoise observed exploitation attempts against the flaw on 24 September, before the vulnerabilities were publicly disclosed, describing attackers attempting to set both the setuid and setgid bits on /bin/sh to obtain a root shell and install a password-protected web shell.5 Palo Alto Networks estimated that roughly 50,000 NetScaler instances remained potentially exposed as of 27 September.5 Mandiant's chief technology officer said "dozens of organisations have been hit, including by suspected state-sponsored threat actors," while researcher Kevin Beaumont separately put the victim count above 100 organisations as of the Tuesday before publication.5 Mandiant also found signs the threat actor may be managing similar web shells across multiple compromised environments, suggesting centralised or coordinated operations, and in at least one case used the SLAPSHOT tunnel to manually explore a victim's internal network and steal credentials.5 No specific threat actor or nation-state has been formally attributed; Mandiant describes the activity only as involving "suspected state-sponsored threat actors" and forecasts broader, opportunistic exploitation by additional actors going forward.5

Update, 29 September, what changed: Mandiant disclosed two previously undocumented malware families deployed against compromised NetScaler appliances: WHIPSHOT, a PHP web shell disguised as a Debian package and stored in the appliance's VPN scripts directory, and SLAPSHOT, a Python-based TCP tunnelling tool that bridges the compromised appliance to internal network devices.4 Mandiant provided new detail on how CVE-2026-88772 is exploited, stating that "transmitting specially malformed or fragmented record headers induces heap memory boundary corruption within the packet engine."4

Attackers modified the appliance's /etc/httpd.conf configuration so .deb package files execute as PHP, and mapped requests for .ico image files under /vpn/media/ to the malicious PHP files, letting the web shells blend in with legitimate NetScaler traffic.4 The web shells use shell_exec() and eval() to run commands, with some returning fake HTTP 404 responses to mask their activity, while attackers set the setuid bit on /bin/sh for persistent root access; SLAPSHOT self-terminates after a period of inactivity to evade detection.4 New indicators of compromise include the presence of /tmp/.uxdport or /tmp/.uxdlock files and suspicious Python processes launched via nohup.4 Mandiant said the attacks began "at least early September," predating public disclosure of the flaws, and have affected government, financial services, education, legal and professional-services organisations across North America and Europe; no victim count was given.4

Update, 28 September, what changed: CISA added CVE-2026-88771 and CVE-2026-88772 to its Known Exploited Vulnerabilities catalogue, stating it had "received reports and partner threat intelligence confirming that threat actors are actively exploiting these vulnerabilities globally."3 The agency also flagged that patching NetScaler devices "can be complex and may require downtime," which prompted the alert to help organisations plan the update.3 Federal civilian agencies must apply the fix by 30 September 2026.3

watchTowr Labs published root-cause analysis on 28 September tracing CVE-2026-88771 to a Perl script, ns_monuploadd_err.pl, that NetScaler uses to process crash and error data.3 The script builds a shell command using attacker-influenced input, so an unauthenticated attacker can inject shell commands by sending a crafted pre-authentication request to the endpoint /nf/auth/doAuthentication.do, with the malicious payload embedded in the login field, achieving remote code execution as root.3

Citrix's published indicators of compromise and incident-response guidance call for organisations to preserve evidence, isolate the device, revoke credentials and access, investigate connected systems, rebuild and update firmware, rotate passwords, keys and certificates, and harden the device, rather than treating this as a routine patch.3

Citrix released security bulletin CTX697096, patching CVE-2026-88771 and CVE-2026-88772, both rated CVSS 9.5.1 CVE-2026-88771 is an improper input validation flaw that lets an unauthenticated attacker run arbitrary commands, and affects all NetScaler ADC and Gateway deployments regardless of configuration.1 CVE-2026-88772 is a memory overflow bug that can cause remote code execution or denial of service on appliances with DTLS enabled, which is on by default for VPN virtual servers.1 Citrix confirmed active exploitation, stating "Exploits of CVE-2026-88771 and CVE-2026-88772 on unmitigated NetScaler deployments have been observed."1

On 26 September, before the patch was released, security firm watchTowr said it was responding to "credible" rumours of unpatched NetScaler remote-code-execution flaws that it had learned of during forensic investigations, and Reddit users reported IT suppliers warning them to shut down NetScaler appliances immediately.1 The Dutch National Cyber Security Centre had separately received a pre-notification about the flaws from an unnamed European partner CERT, and Citrix reportedly filed a notice under the EU's Cyber Resilience Act after detecting the attacks.2 The Dutch agency said exploitation had reportedly hit "multiple Citrix customers worldwide."2

The patch bundle also fixes six additional, unexploited vulnerabilities, and applies to NetScaler ADC and Gateway 14.1 before 14.1-73.37 and 13.1 before 13.1-64.23, along with the FIPS and NDcPP variants.1 Builds that already patched August's CVE-2026-19490 authentication-bypass flaw remain in the vulnerable range and need this update too, and the 13.1 branch received a fix despite reaching end of maintenance on 15 September.1

Why it matters: a third emergency NetScaler patch inside two weeks, this time for a flaw Citrix says it hasn't seen exploited yet, shows the vendor is now racing ahead of attackers rather than reacting to confirmed intrusions, but with recent NetScaler proof-of-concepts turning into mass exploitation within hours of going public, administrators who skip this one on the strength of "not yet exploited" are gambling against a very short clock.

Sources

  1. Warning: Two Unpatched Citrix NetScaler RCE Zero-Days Under Active Exploitation (The Hacker News)
  2. Citrix admins warned to shut down NetScalers over 2 exploited zero-days (BleepingComputer)
  3. CISA Says Attackers Are Exploiting Two Critical Citrix NetScaler Flaws Globally (The Hacker News)
  4. Hackers exploit Citrix NetScaler zero-day to deploy web shells (BleepingComputer)
  5. Government, Finance Orgs Targeted in Weeks-Long NetScaler Zero-Day Attacks (SecurityWeek)
  6. Citrix NetScaler Post-Exploitation Payload Creates Superuser, Maps Web Shell to CSS-Like URLs (The Hacker News)
  7. Netscaler admins beware: Zero-day causes crashes and code execution (heise online)
  8. NetScaler CVE-2026-88771: patching leaves the web shell (hard2bit)
  9. Citrix patches NetScaler SAML zero-day exploited in attacks (BleepingComputer)
  10. New Citrix NetScaler Issue Emerges (Decipher)
  11. More Shells Than a Seafood Buffet: Tracking Citrix NetScaler Exploitation Activities (CVE-2026-88771) (eSentire)
  12. Four Attacker Clusters Exploit Citrix NetScaler Flaw CVE-2026-88771 to Plant Web Shells (SecurityOnline.info)
  13. Citrix Ships Another Emergency Fix for NetScaler Appliances (BankInfoSecurity)
  14. Critical Remote Code Execution Vulnerability in Citrix NetScaler (NHS England Digital)