Someone left the master key to every VM taped to the syslog socket
CVE-2026-59310 is a directory traversal vulnerability in VMware vCenter Server's built-in Syslog service. An unauthenticated attacker with network access to the Syslog listener (ports 514/TCP, 1514/TCP, 514/UDP) can craft path-traversal payloads that escape the intended directory and achieve arbitrary code execution on the vCenter appliance as root. Affected versions span vCenter 8.0 (all builds before U3k), vCenter 9.0.x (before 9.0.2.0100), and vCenter 9.1.x (before 9.1.0.0300), plus downstream products VMware Cloud Foundation 5.x/9.x, vSphere Foundation 9.x, and Telco Cloud Platform 3.0–5.1.x. Broadcom states no workarounds exist.
Broadcom rates this CRITICAL at 9.8 and the vendor severity is fully justified. vCenter is not a random appliance — it *is* the management plane for every ESXi host and every VM in your environment. The CVSS vector (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) correctly captures the no-auth, no-interaction, full-impact nature of the bug. While vCenter should never be internet-facing, it is always reachable on management VLANs, meaning any attacker with internal network access — post-phish, post-VPN, post-contractor-laptop — can reach it. The 9.8 is not hype; it reflects a genuinely catastrophic attack surface on a component whose compromise cascades to the entire virtualized estate.
4 steps from start to impact.
Reach the vCenter Syslog listener
- Network path to vCenter Syslog port (514 or 1514)
- vCenter should be on an isolated management VLAN — but many orgs have flat networks or overly permissive firewall rules
- Internet-facing exposure is low (~6,700–14,000 instances historically) compared to total install base
vmware on port 443 reveal externally exposed instances; internal asset inventories should flag vCenter IPsSend crafted syslog message with path traversal payload
../../../) to the Syslog service. No authentication is required — syslog on ports 514/1514 has no auth mechanism by design. The traversal escapes the intended log directory and allows writing to arbitrary filesystem locations on the vCenter appliance.- No authentication needed
- Syslog service is enabled by default on vCenter
- No known public PoC exists yet — the attacker needs to reverse-engineer the patch diff or develop the payload independently
../ sequences; deep packet inspection on syslog trafficAchieve arbitrary file write → code execution
- Write target must be a location that triggers execution (cron, systemd, web shell path)
- Appliance may have SELinux or AppArmor policies limiting writable paths — but VCSA historically has permissive configs
Pivot to ESXi hosts and VMs
- vCenter manages ESXi hosts (this is its entire purpose)
- None — this is the expected operational model of vCenter
The supporting signals.
| In-the-Wild Exploitation | Not observed. No evidence of active exploitation as of 2026-07-31. Not yet on CISA KEV. However, vCenter has appeared on KEV ten times historically — exploitation is a matter of *when*, not *if*. |
|---|---|
| Proof-of-Concept | No public PoC available as of 2026-07-31. Discovered by Phil Brass and Matt South of Atredis Partners. Patch diffing is expected to yield weaponizable details within days to weeks. |
| EPSS Score | Not yet scored (disclosed 2026-07-30). Historical vCenter RCE CVEs have reached top-1% EPSS within weeks of disclosure. |
| KEV Status | Not listed as of 2026-07-31. Given vCenter's KEV history (10 prior entries), addition is probable if exploitation is observed. |
| CVSS Vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H — network-accessible, no auth, no interaction, full CIA impact. This is the worst possible base score for Scope:Unchanged. |
| Affected Versions | vCenter 8.0 (all builds), 9.0.x, 9.1.x; VMware Cloud Foundation 5.x, 9.0.x, 9.1.x; vSphere Foundation 9.0.x, 9.1.x; Telco Cloud Platform 3.0, 4.x, 5.0.x, 5.1.x; Telco Cloud Infrastructure 3.0 |
| Fixed Versions | vCenter 8.0 U3k, 9.0.2.0100, 9.1.0.0300; async patches via KB88287 and KB449886 |
| Exposure Data | Shodan/Censys historically index 6,700–14,000 internet-facing vCenter instances. Internal exposure is universal — every enterprise with VMware has vCenter on the management network. |
| Workarounds | None available. Broadcom explicitly states no workaround exists. Patching or network isolation are the only options. |
| Advisory | VMSA-2026-0006 — published 2026-07-29 |
noisgate verdict.
vCenter Server is *the* hypervisor management plane — 100% of its installs occupy the highest-value infrastructure role by definition, and RCE on it cascades to every ESXi host and VM in the environment. The single most decisive factor is role multiplier: unauth RCE on the canonical fleet-control-plane component means compromise of the entire virtualized estate.
Why this verdict
- Unauth network RCE with zero interaction — the CVSS vector AV:N/AC:L/PR:N/UI:N is the most dangerous attack profile possible. No credentials, no phishing, no user clicks required.
- Role multiplier: vCenter IS the hypervisor control plane. This is not a component that *sometimes* runs in a high-value role — it exists exclusively to manage ESXi hosts and VMs. Compromise of vCenter gives the attacker root on vCenter → credentials for all ESXi hosts → control of every VM → domain controllers, databases, backup servers, the entire fleet. Blast radius is fleet-scale by definition. This alone floors the verdict at CRITICAL.
- No workarounds exist. Broadcom explicitly confirms there is no mitigation short of patching or total network isolation of the Syslog ports. You cannot disable the Syslog service without breaking log forwarding functionality.
- vCenter's KEV track record is 10 prior entries. History shows that vCenter RCE bugs are weaponized rapidly (CVE-2021-21985, CVE-2023-34048, CVE-2024-37079). The window between disclosure and exploitation has been shrinking with each iteration.
- Management network reachability is not meaningful friction. While vCenter is not internet-facing in best-practice deployments, it is always reachable from the management VLAN. Any attacker with internal network access — post-phish, compromised VPN, rogue contractor, or lateral movement from any server subnet — can reach it. 'Requires internal network' is not a downgrade factor for a post-initial-access scenario.
Why not higher?
A 9.8 is already the maximum practical CVSS score for Scope:Unchanged. There is no higher severity bucket than CRITICAL. The only thing preventing a 10.0 is that Scope is Unchanged in the CVSS vector (the Syslog service itself is the vulnerable component, not a cross-scope boundary like a VM escape).
Why not lower?
Downgrading below CRITICAL would require evidence that the attack chain has meaningful friction or that vCenter occupies a low-value role in some deployments — neither is true. The bug is unauthenticated, requires no interaction, has no workaround, and the target component is the single most important infrastructure management server in any VMware environment. Even the absence of a public PoC is temporary — Atredis Partners' responsible disclosure and the patch diff will yield exploitation details rapidly. Historical precedent (CVE-2021-21985 weaponized in <2 weeks) confirms this.
What to do — in priority order.
- Firewall the vCenter Syslog ports immediately — Block inbound traffic to ports 514/TCP, 514/UDP, and 1514/TCP on every vCenter appliance from all sources except explicitly authorized syslog *clients* (ESXi hosts that forward logs to vCenter). This is your only compensating control since no vendor workaround exists. Deploy within the noisgate mitigation SLA of 3 days for CRITICAL findings. Use host-based firewall rules (iptables on VCSA) and network ACLs on the management VLAN switch/firewall.
- Restrict management VLAN access — Tighten ACLs on your management network segment so only jump hosts and authorized admin workstations can reach vCenter on *any* port. This reduces the population of hosts that could be used as a pivot point. If you don't already have microsegmentation here, this CVE is your forcing function.
- Enable FIM on vCenter appliance — Deploy or verify file integrity monitoring on the VCSA appliance filesystem, focusing on
/etc/cron.d/,/etc/systemd/system/,/usr/lib/vmware-syslog/, and web shell paths under/usr/lib/vmware-vsphere-ui/. This provides detection-in-depth if the compensating firewall rule is bypassed. - Monitor vSphere API for anomalous operations — Alert on unusual vSphere API calls — bulk VM snapshot creation, VIB installation on ESXi hosts, SSO credential extraction, or permission changes — which would indicate post-exploitation activity if the vulnerability is exploited before patching.
- WAF/reverse proxy in front of vCenter — this vulnerability targets the Syslog service (port 514/1514), not the HTTPS management interface (port 443). A WAF only inspects HTTP/HTTPS traffic and provides zero coverage for syslog protocol exploitation.
- Disabling the vCenter Syslog service — while theoretically removing the attack surface, this breaks ESXi host log forwarding and compliance logging. Broadcom does not document this as a supported configuration change, and it may cause VCSA health check failures.
- Network-level IDS/IPS with generic syslog signatures — standard syslog inspection rules look for format violations, not directory traversal payloads embedded in message fields. Until vendors release specific signatures for CVE-2026-59310, generic detection is unreliable.
Crowdsourced verification payload.
Run this script on each vCenter Server Appliance (VCSA) via SSH as root. Example: ssh [email protected] 'bash -s' < check_cve_2026_59310.sh. Requires root shell access to the VCSA.
#!/bin/bash
# check_cve_2026_59310.sh
# Checks VMware vCenter Server for CVE-2026-59310 (Syslog directory traversal)
# Run on the VCSA appliance as root.
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN
set -euo pipefail
PATCHED_BUILDS_8="25163495" # 8.0 U3k build number (verify against release notes)
PATCHED_VER_90="9.0.2.0100"
PATCHED_VER_91="9.1.0.0300"
# Get vCenter version and build
if command -v vpxd >/dev/null 2>&1; then
FULL_VER=$(vpxd -v 2>/dev/null || echo "")
elif [ -f /etc/vmware/version ]; then
FULL_VER=$(cat /etc/vmware/version)
else
echo "UNKNOWN - Cannot determine vCenter version. Is this a VCSA?"
exit 2
fi
if [ -z "$FULL_VER" ]; then
echo "UNKNOWN - Could not read vCenter version info."
exit 2
fi
echo "Detected vCenter: $FULL_VER"
# Extract version components
VER=$(echo "$FULL_VER" | grep -oP '[0-9]+\.[0-9]+\.[0-9]+' | head -1 || echo "")
BUILD=$(echo "$FULL_VER" | grep -oP 'build-\K[0-9]+' || echo "")
if [ -z "$VER" ]; then
echo "UNKNOWN - Could not parse version string from: $FULL_VER"
exit 2
fi
MAJOR=$(echo "$VER" | cut -d. -f1)
MINOR=$(echo "$VER" | cut -d. -f2)
echo "Parsed version: $VER (build: ${BUILD:-unknown})"
# Version comparison function
ver_gte() {
printf '%s\n%s' "$2" "$1" | sort -V -C
}
if [ "$MAJOR" -eq 8 ]; then
# vCenter 8.0 — check for U3k build
if [ -n "$BUILD" ] && [ "$BUILD" -ge "$PATCHED_BUILDS_8" ] 2>/dev/null; then
echo "PATCHED - vCenter 8.0 build $BUILD is >= U3k ($PATCHED_BUILDS_8)"
exit 0
else
echo "VULNERABLE - vCenter 8.0 build ${BUILD:-unknown} is below U3k ($PATCHED_BUILDS_8)"
exit 1
fi
elif [ "$MAJOR" -eq 9 ] && [ "$MINOR" -eq 0 ]; then
if ver_gte "$VER" "9.0.2"; then
echo "PATCHED - vCenter $VER meets or exceeds fixed version $PATCHED_VER_90"
exit 0
else
echo "VULNERABLE - vCenter $VER is below fixed version $PATCHED_VER_90"
exit 1
fi
elif [ "$MAJOR" -eq 9 ] && [ "$MINOR" -eq 1 ]; then
if ver_gte "$VER" "9.1.0"; then
echo "PATCHED - vCenter $VER meets or exceeds fixed version $PATCHED_VER_91"
exit 0
else
echo "VULNERABLE - vCenter $VER is below fixed version $PATCHED_VER_91"
exit 1
fi
else
echo "UNKNOWN - vCenter version $VER is outside known affected ranges (8.0, 9.0.x, 9.1.x)"
exit 2
fiIf you remember one thing.
Sources
- VMSA-2026-0006 — Broadcom Security Advisory
- Rapid7 ETR — CVE-2026-59309 & CVE-2026-59310 Analysis
- BleepingComputer — VMware fixes three critical flaws
- GBHackers — Critical VMware vCenter Flaws
- Qualys ThreatPROTECT — VMware ESX, vCenter Multiple Vulnerabilities
- Field Effect — Broadcom Patches Critical VMware Vulnerabilities
- NVD — CVE-2026-59310
- HADESS — vCenter Attack Surface Analysis
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.