Two 9.1s in Zscaler Client Connector: unauth RCE and portal auth bypass
ZCC has a pair of critical bugs (CVE-2026-59568, CVE-2026-59564, both CVSS 9.1) that could hand attackers full endpoint control or let them hijack agent-to-portal trust. Also: AWX path traversal via archive extraction (7.2), WebKitGTK use-after-free (8.8), and a CRLF injection in TIM Flow (8.1). None exploited in the wild yet.
Two CVSS 9.1 bugs in Zscaler Client Connector landed today, and both are bad. CVE-2026-59568 lets an attacker run code without authentication on any endpoint running ZCC, which usually means SYSTEM-level access across your whole fleet. CVE-2026-59564 breaks the trust between ZCC and its management portal, opening the door to policy manipulation and traffic interception. Neither is exploited in the wild yet, but if you're a Zscaler shop, don't wait on these.
Today's CVEs
Sorted by urgencyCVE-2026-59568
NVDAn attacker can run arbitrary code through Zscaler Client Connector without authentication or any special privileges. CVSS 9.1, and since ZCC usually runs with elevated permissions on endpoints, a successful exploit could give an attacker full control of the machine. Not yet exploited in the wild, but the attack surface here is wide if you have ZCC deployed across your fleet.
- Affected estate
- Anyone running Zscaler Client Connector on endpoints
- How to check
- Open ZCC on an endpoint and check the About/version info, or query your endpoint management tool for installed ZCC versions. Compare against the fixed version listed in Zscaler's advisory.
- Included because
- unauthenticated; remote code execution; CVSS 9.1; widely deployed endpoint agent
- Action
- Push the updated ZCC package through the Zscaler admin portal or your software deployment pipeline.
- Why it matters
- Unauthenticated RCE in an agent that runs on every endpoint means an attacker could compromise machines across your entire fleet.
- Source
- Zscaler advisory
Evidence trail
- NVD: View source
CVE-2026-59564
NVDAn attacker can bypass authentication between the Zscaler Client Connector and its management portal. That could let someone impersonate the agent, manipulate policy enforcement, or intercept traffic routing decisions. CVSS 9.1, no exploitation in the wild yet, but the trust relationship between ZCC and its portal is a critical control point.
- Affected estate
- Anyone running Zscaler Client Connector and using the ZCC Portal for management
- How to check
- Verify the installed ZCC version on endpoints and confirm the portal version in the Zscaler admin console.
- Included because
- unauthenticated; authentication bypass; CVSS 9.1; affects trust boundary between agent and management plane
- Action
- Upgrade ZCC on all endpoints and confirm the portal component is patched.
- Why it matters
- An authentication bypass on the agent-to-portal channel could let an attacker subvert your zero-trust enforcement entirely.
- Source
- Zscaler advisory
Evidence trail
- NVD: View source
CVE-2026-71364
NVDAWX's project archive extraction doesn't sanitize filenames, so a malicious zip or tar archive can write files anywhere on the execution node's filesystem. If an attacker controls the archive content (compromised SCM, MITM on HTTP, or a malicious URL), they can drop cron jobs, SSH keys, or poisoned playbooks to get remote code execution. This requires the attacker to control or tamper with the archive source, so it's not a zero-click from the internet, but any AWX setup pulling projects over plain HTTP or from untrusted sources is at real risk.
- Affected estate
- Teams running AWX (the open-source Ansible Tower upstream) for automation
- How to check
- Run 'awx --version' or check the AWX web UI's About page. Review project sources for any using plain HTTP URLs.
- Included because
- path traversal to arbitrary file write; potential RCE; automation infrastructure is high-value target; CVSS 7.2
- Action
- Upgrade AWX to the patched release. Audit project sources and enforce HTTPS for all archive URLs.
- Why it matters
- Arbitrary file write on your automation controller can turn into full RCE, and your AWX nodes likely have SSH access to your entire managed fleet.
- Source
- AWX project advisory
Evidence trail
- NVD: View source
CVE-2026-78376
NVDA use-after-free bug in WebKitGTK lets an attacker run code on your system by getting a user to visit a malicious web page. CVSS 8.8. This affects any Linux application that uses WebKitGTK for rendering web content, including GNOME Web (Epiphany) and other GTK-based browsers or embedded webviews. It does require user interaction: someone has to load the malicious content.
- Affected estate
- Linux sysadmins running desktops or applications that use WebKitGTK (GNOME, Epiphany, Evolution, GIMP help browser, etc.)
- How to check
- Run 'rpm -q webkit2gtk3' (RHEL/Fedora) or 'dpkg -l libwebkit2gtk*' (Debian/Ubuntu) to check the installed version.
- Included because
- memory corruption; user interaction required but low complexity; CVSS 8.8; common Linux component
- Action
- Apply the latest webkit2gtk3 or libwebkit2gtk update from your distro's repos.
- Why it matters
- Memory corruption via web content can give an attacker code execution under the user's context on any affected Linux workstation.
- Source
- WebKitGTK advisory / distro security tracker
Evidence trail
- NVD: View source
CVE-2026-39915
NVDTIM Flow before 26.0.6 lets an attacker inject CRLF sequences into the 'rt' URL parameter, which gets reflected into Set-Cookie headers. That means an attacker can craft a link that, when clicked by an authenticated user, runs JavaScript in their browser session. The practical result: session hijacking and account credential changes. CVSS 8.1, but it does require a user to click a crafted link.
- Affected estate
- Teams running TIM Flow versions before 26.0.6
- How to check
- Check the TIM Flow version in the application's admin console or deployment configuration.
- Included because
- CRLF injection; reflected into headers; session hijack risk; CVSS 8.1; user interaction required
- Action
- Upgrade to TIM Flow 26.0.6 or later.
- Why it matters
- CRLF injection into response headers enables session theft and credential changes for any authenticated user who clicks a malicious link.
- Source
- TIM Flow vendor advisory
Evidence trail
- NVD: View source
One email, every Wednesday morning.
SubscribeFrom the field notes
From this beat
Read the rest of the field notes โ