A locally-signed-in agent can forge messages to the Remote Control helper and ride it up to whatever privileges that helper runs as
CVE-2026-53406 is a CWE-345 (Insufficient Verification of Data Authenticity) flaw in the Remote Control component shipped with Zoom Contact Center for Windows, versions prior to 7.0.0. The component accepts control messages without properly validating their origin, so a process running as a low-privilege local user on the same Windows host can craft messages that the helper trusts and acts on. Net effect: local privilege escalation on a machine where the attacker already has an authenticated user session.
Vendor's HIGH (7.8) rating leans on the CIA-triple impact (C:H/I:H/A:H) of a successful EoP. That math is technically correct but ignores who actually runs the Remote Control for Zoom Contact Center binary — it's a niche add-on for contact-center agents, not the general Zoom Workplace install base. Combined with AV:L (no remote attack surface) and PR:L (you must already be a logged-in user), the realistic reading is MEDIUM. There's no KEV listing, EPSS is 0.08%, and no public PoC has surfaced.
4 steps from start to impact.
Attacker gets an authenticated local session
- Valid local user credentials or an existing low-priv foothold
- Target host has Remote Control for Zoom Contact Center < 7.0.0 installed
- Component service/helper is running
- Contact-center agents are typically the only population with this component installed
- EDR/AV usually flags unsigned local binary drops before this stage
- Most enterprise builds restrict interactive logon to specific user groups
Locate and reach the Remote Control IPC channel
PipeList, Process Explorer, RpcView, or a one-off WinDbg attach. No weaponized public tool is specific to this CVE yet.- Local user can open the IPC endpoint's ACL
- Helper accepts cross-session or cross-integrity-level connections
- Windows session isolation may block cross-session pipe access depending on how the helper was configured
- If the helper runs at Medium integrity, the escalation ceiling is low
Forge a control message the helper trusts
- Knowledge of the message format (reverse-engineered from the binary)
- Helper path that maps a forged message to a privileged action
- Reverse engineering effort is non-trivial and bug-specific
- Some forged-message paths only yield read primitives, not write/exec
Privileged action executes as the helper context
- Helper service token is higher-privileged than the attacker's token
- Forged message maps to an exec/write primitive
- Blast radius is local — no lateral movement primitive in the bug itself
- Single-host compromise still requires AD/credential-theft step to scale
The supporting signals.
| In-the-wild status | No active exploitation observed. Not on CISA KEV. |
|---|---|
| Public PoC | None published as of 2026-06-30. No GitHub repos, no Metasploit module. |
| EPSS | 0.00081 (≈ 0.08%) — bottom percentile, exploitation unlikely in 30 days |
| KEV status | Not listed |
| CVSS vector | CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H — local, low-priv prerequisite, full CIA on success |
| Affected versions | Remote Control for Zoom Contact Center for Windows < 7.0.0 |
| Fixed version | 7.0.0 and later (Zoom advisory ZSB-26xxx) |
| Exposure data | No external attack surface — AV:L. Shodan/Censys irrelevant. Install base limited to Zoom Contact Center agent population. |
| Disclosure date | 2026-06-12 via Zoom Security Bulletin |
| Reporter | Zoom Offensive Security Team (internal discovery, per bulletin pattern) |
noisgate verdict.
Downgraded because the decisive factor is AV:L + PR:L — the attacker must already hold a valid authenticated local session on a host running a niche Zoom add-on, which means the bug only matters *after* an initial-access stage has already succeeded. Installed base is bounded to contact-center agent workstations, not the broader Zoom Workplace fleet, so blast-radius math is single-host EoP with no lateral primitive.
Why this verdict
- Authentication required (PR:L): the attacker is already past initial access — this is a post-compromise amplifier, not an entry point.
- No network reachability (AV:L): zero exposure to the internet or LAN; no remote-trigger primitive exists.
- Narrow install base: Remote Control for Zoom Contact Center ships only with the Contact Center add-on, not core Zoom Workplace — most Zoom-using enterprises are not affected at all.
- Role multiplier — low-value role (contact-center agent workstation): chain succeeds, but blast radius is single-host EoP. No DC, no hypervisor, no CI/CD, no identity-plane role plausibly runs this component.
- Role multiplier — typical role (shared agent kiosk / VDI): chain still succeeds and yields SYSTEM-on-the-kiosk; still bounded to one host. Lateral movement requires a separate credential-theft step.
- No high-value role exists for this component — it is not canonically deployed on identity, PKI, hypervisor, backup, or kernel-mode-agent hosts, so the HARD-RULE floor for HIGH does not engage.
- Threat-intel quiet: EPSS 0.08%, no PoC, no KEV, no campaign chatter — opportunistic exploitation pressure is near zero.
Why not higher?
HIGH (vendor's call) requires either a remote vector, a high-value-role multiplier, or KEV/active-exploitation pressure — none of which apply. The component is not deployed on identity-plane or kernel-mode-agent hosts, so the deployment-role floor does not push the verdict up.
Why not lower?
LOW would undersell a confirmed C:H/I:H/A:H EoP primitive that lands SYSTEM (or elevated) on a corporate endpoint. EoP-to-SYSTEM is still a meaningful step in an attacker's kill chain even if it requires a prior foothold, so MEDIUM is the honest read.
What to do — in priority order.
- Inventory which Windows endpoints actually have Remote Control for Zoom Contact Center installed — Most orgs running Zoom do *not* have the Contact Center add-on — scope first. Use SCCM/Intune/Tanium to query for the helper binary path or installed-product GUID. Do this within the noisgate remediation window — there is no mitigation SLA at MEDIUM.
- Force-upgrade affected agents to Remote Control for Zoom Contact Center ≥ 7.0.0 — Push via your existing Zoom update channel or MSI deployment. Schedule alongside your normal monthly patch ring; no need to break change windows.
- Restrict interactive logon on contact-center workstations to the agent role only — Removes the casual-attacker population. Group Policy 'Allow log on locally' / 'Deny log on locally' scoped to the contact-center OU shrinks the prerequisite population for *every* local-auth bug, not just this one.
- Enable EDR child-process rules for the Zoom Remote Control helper — Alert on cmd.exe / powershell.exe / rundll32.exe spawning from the helper's process tree. Catches the post-EoP execution stage even without a signature for the forged message.
- Network segmentation / firewall ACLs — bug is AV:L, no network packet to block.
- Disabling SMB / NTLM hardening — unrelated Windows-auth class; this is local IPC, not network auth.
- WAF / reverse proxy rules in front of Zoom — Zoom traffic is outbound to Zoom cloud; no inbound surface to filter.
- MFA on Zoom accounts — protects sign-in, doesn't change a local Windows-session attacker's reach to the local helper.
Crowdsourced verification payload.
Run on each Windows endpoint suspected to have Zoom Contact Center installed (i.e., contact-center agent workstations). Invoke from an admin PowerShell session: powershell -ExecutionPolicy Bypass -File .\Check-CVE-2026-53406.ps1. Requires local admin to read installed-product registry hives reliably; outputs VULNERABLE / PATCHED / UNKNOWN.
# Check-CVE-2026-53406.ps1
# Detects vulnerable Remote Control for Zoom Contact Center (< 7.0.0)
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN / not installed
$ErrorActionPreference = 'SilentlyContinue'
$fixedVersion = [Version]'7.0.0'
$found = $false
$vuln = $false
$observed = $null
# 1. Look in 32/64-bit Uninstall registry hives
$uninstallPaths = @(
'HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*',
'HKLM:\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*'
)
foreach ($p in $uninstallPaths) {
Get-ItemProperty $p | Where-Object {
$_.DisplayName -match 'Remote Control.*Zoom Contact Center' -or
$_.DisplayName -match 'Zoom Contact Center.*Remote Control'
} | ForEach-Object {
$found = $true
$observed = $_.DisplayVersion
try {
if ([Version]$_.DisplayVersion -lt $fixedVersion) { $vuln = $true }
} catch { $vuln = $true } # unparseable version = treat as vulnerable
}
}
# 2. Fallback: probe common install paths for the helper binary
if (-not $found) {
$candidatePaths = @(
"$env:ProgramFiles\Zoom\ContactCenter\RemoteControl",
"${env:ProgramFiles(x86)}\Zoom\ContactCenter\RemoteControl",
"$env:LOCALAPPDATA\Zoom\ContactCenter\RemoteControl"
)
foreach ($path in $candidatePaths) {
if (Test-Path $path) {
$exe = Get-ChildItem $path -Filter '*.exe' -Recurse | Select-Object -First 1
if ($exe) {
$found = $true
$observed = $exe.VersionInfo.FileVersion
try {
if ([Version]$exe.VersionInfo.FileVersion -lt $fixedVersion) { $vuln = $true }
} catch { $vuln = $true }
}
}
}
}
if (-not $found) {
Write-Output 'UNKNOWN: Remote Control for Zoom Contact Center not detected on this host'
exit 2
}
if ($vuln) {
Write-Output "VULNERABLE: detected version $observed (< 7.0.0)"
exit 1
}
Write-Output "PATCHED: detected version $observed"
exit 0
If you remember one thing.
Sources
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.