← Back to Feed CACHED · 2026-07-10 09:34:06 · CACHE_KEY tenable:87942
tenable:87942 · CWE-284 · Disclosed 2016-01-07

ESXi 5

ASSESSED — NOISGATE V0.5
Vendor
Reassessed
Verdict:
Do you agree?
01 · The Real Story

A stale memory-corruption bug in a feature nobody enables, on a hypervisor version nobody should still be running

CVE-2015-6933 is a kernel memory corruption bug in the VMware Tools HGFS (Host-Guest File System) Shared Folders driver on Windows guests. A local, low-privileged user *inside the Windows VM* can corrupt guest kernel memory via crafted HGFS traffic and escalate to SYSTEM within that same guest. The affected fleet is ESXi 5.5 hosts prior to build 3248547 (patch ESXi550-201512102-SG, Dec 2015), plus companion updates to Workstation 11.x <11.1.2, Player 7.x <7.1.2, Fusion 7.x <7.1.2, and ESXi 5.0/5.1/6.0 lines with their own builds.

Tenable/VMware rated this MEDIUM (CVSS 6.3, AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L) and that label overstates the practical risk. The scope is explicitly unchanged — VMware itself notes host memory cannot be manipulated from the guest, so this is *not* a VM escape. It is a guest-local privilege escalation that only fires when Shared Folders is enabled (rare on server workloads, disabled by default on ESXi-hosted Windows guests) and requires an attacker already executing code inside the VM.

"Guest-to-guest LPE inside a Windows VM on a decade-old ESXi 5.5 host that Broadcom stopped supporting in 2020. Backlog it."
02 · The Attack Path

4 steps from start to impact.

STEP 01

Land code execution inside a Windows guest VM

Attacker must first obtain a low-privileged interactive or remote foothold on a Windows guest running on the vulnerable ESXi 5.5 host. Any standard initial-access vector works — phish, stolen creds, exposed RDP — but the vuln itself contributes nothing here.
Conditions required:
  • Windows guest OS on ESXi 5.5 <3248547
  • VMware Tools installed in the guest
  • Local user shell (any privilege level)
Where this breaks in practice:
  • Requires prior compromise of the guest — this is not a remote entry point
  • Server VMs rarely have interactive users to phish
Detection/coverage: Standard EDR on the guest catches the initial-access tradecraft; the HGFS bug is invisible until step 3.
STEP 02

Confirm Shared Folders (HGFS) is enabled

The vulnerable code path lives in the vmhgfs.sys guest driver, reachable only when Shared Folders is configured for the VM. On ESXi (as opposed to Workstation/Fusion), Shared Folders is disabled by default and is not a supported feature in most server deployments.
Conditions required:
  • HGFS driver loaded (vmhgfs.sys present)
  • Shared Folders enabled at VMX/vmtoolsd layer
Where this breaks in practice:
  • Feature is typically absent on production server guests
  • Many hardened images strip vmhgfs.sys via VMware Tools custom install
Detection/coverage: driverquery / Autoruns will show vmhgfs.sys; PowerShell Get-Service VMTools plus checking .vmx for sharedFolder entries.
STEP 03

Trigger HGFS memory corruption via crafted RPC

Attacker sends malformed HGFS requests through the \\.\HGFS device (or the backdoor RPC channel) that fail input validation and corrupt kernel pool memory. Public PoC details remain sparse — the bug was disclosed in VMSA-2016-0001 without a weaponized exploit released, and no Metasploit module exists.
Conditions required:
  • Ability to open HGFS device handle from user mode
  • Reliable heap-grooming primitive in guest kernel
Where this breaks in practice:
  • Windows kernel mitigations (kASLR, SMEP on modern guests) complicate reliable EoP
  • No public weaponized PoC — attacker must write their own
Detection/coverage: EDR kernel telemetry may catch the resulting pool corruption crash; BSODs with vmhgfs.sys in the stack are the signature.
STEP 04

Escalate to SYSTEM inside the guest

Successful corruption yields arbitrary kernel write / token replacement, giving the attacker NT AUTHORITY\SYSTEM in that single guest. Scope is Unchanged — the hypervisor is not touched, other guests are not affected, and this is not a VM escape.
Conditions required:
  • Reliable exploitation primitive from step 3
Where this breaks in practice:
  • Blast radius is one Windows VM
  • Any other Windows LPE from the last decade achieves the same outcome with less engineering
Detection/coverage: Post-exploit SYSTEM-level activity is the primary detection surface, not the exploit itself.
03 · Intelligence Metadata

The supporting signals.

CVECVE-2015-6933
Vendor advisoryVMSA-2016-0001 (VMware, 2016-01-07)
CVSS v3.06.3 MEDIUMAV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:L (scope Unchanged — guest-local, not VM escape)
CVSS v26.5 — AV:N/AC:L/Au:S/C:P/I:P/A:P
CWECWE-284 Improper Access Control (memory corruption via unvalidated HGFS input)
KEV statusNot listed in CISA KEV as of 2026-07-10
EPSSSub-1% (<0.5 percentile) — no observed exploitation telemetry
Public PoCNone weaponized publicly; no Metasploit/ExploitDB entry; researcher Tavis Ormandy (Google Project Zero) credited alongside coordinated report
AffectedESXi 5.5 <build 3248547; also ESXi 5.0/5.1/6.0, Workstation 11.x <11.1.2, Player 7.x <7.1.2, Fusion 7.x <7.1.2 — only when Windows guest runs vulnerable VMware Tools *and* Shared Folders is enabled
Fixed inESXi 5.5 build 3248547 via patch ESXi550-201512102-SG (Dec 2015). Updating VMware Tools in Windows guests is required for full mitigation.
LifecycleESXi 5.5 General Support ended 2018-09-19, Technical Guidance ended 2020-09-19. Broadcom does not ship fixes for this line anymore.
ExposureShodan/Censys show a dwindling long tail of ESXi 5.5 hosts, mostly in unmanaged SMB/lab environments; no scanner reports mass targeting of HGFS.
04 · The Call

noisgate verdict.

Final Verdict
DOWNGRADED to LOW (3.1/10)

Decisive factor: this is a guest-local privilege escalation with scope Unchanged — no VM escape, no cross-tenant impact, blast radius capped at one Windows guest that the attacker already has code execution on. Combined with a rare-by-default feature (HGFS Shared Folders) and a hypervisor version that exited vendor support in 2020, the practical risk sits well below the MEDIUM CVSS label.

HIGH Scope and blast radius (guest-local, not hypervisor escape)
HIGH Absence of active exploitation and KEV listing
MEDIUM Real-world prevalence of Shared Folders on server workloads

Why this verdict

  • Scope Unchanged (S:U): VMware explicitly states host memory cannot be manipulated from the guest — this is guest-internal EoP, not a hypervisor breakout. That collapses the multiplier that would justify MEDIUM+ on a hypervisor CVE.
  • Prerequisite stack is deep: attacker needs (a) code execution inside a Windows guest, (b) Shared Folders enabled on that guest, and (c) a working exploit for a bug with no public weaponized PoC. Each condition compounds downward.
  • Role multiplier — hypervisor host: the chain does NOT succeed as a hypervisor compromise; the bug lives in the guest driver, not the VMkernel. The blast radius on ESXi is *one guest VM*, not the fleet. This is the opposite of the ESXi CVEs that earn a CRITICAL floor.
  • Role multiplier — Windows guest: even in the worst case the attacker gets SYSTEM on one VM they already control. Any modern Windows LPE (PrintNightmare, CLFS, etc.) delivers the same outcome with orders of magnitude less effort.
  • No exploitation signal: not in KEV, EPSS trivial, no Metasploit module, 10+ years without a public campaign leveraging this vector.
  • Feature disabled by default on ESXi: the exposed installed-base share where HGFS is actively used on a server workload is a small fraction of an already-shrinking ESXi 5.5 population.

Why not higher?

MEDIUM would imply the bug meaningfully changes an attacker's position on a typical ESXi 5.5 estate. It doesn't — the attacker must already be inside a guest, the exploited feature is off by default on servers, and success yields SYSTEM on that same guest only. There is no hypervisor impact, no lateral primitive across VMs, and no evidence of in-the-wild use in a decade.

Why not lower?

IGNORE would be wrong because ESXi 5.5 is out of support and any residual host running vulnerable Tools represents a genuine (if narrow) EoP path. LOW captures the reality: log it, fold it into the ESXi 5.5 decommission project, but do not burn a change window on it in isolation.

05 · Compensating Control

What to do — in priority order.

  1. Decommission ESXi 5.5 outright — ESXi 5.5 exited Technical Guidance on 2020-09-19. Any host still running it is carrying a long tail of unpatched vulns far scarier than this one. Fold this CVE into the migration to ESXi 7.0U3 or 8.x and close the whole class of issues at once — target completion within the noisgate remediation SLA for LOW (365-day backlog window).
  2. Disable Shared Folders on all Windows guests — In the VM .vmx, set isolation.tools.hgfs.disable = "TRUE" and restart the guest. This kills the HGFS RPC channel entirely and neuters the vulnerable code path without touching the host. Push via PowerCLI across the cluster in one change window.
  3. Remove the Shared Folders component from VMware Tools — Reinstall VMware Tools in each Windows guest with a Custom install and deselect the *Shared Folders* feature so vmhgfs.sys is never loaded. This is the belt-and-suspenders complement to the VMX flag.
  4. Enforce least privilege inside guest VMs — Because this bug requires a local user in the guest, removing interactive logon for service accounts, enforcing LAPS on guest local admins, and restricting RDP reduces the population of accounts that could ever reach step 1.
What doesn't work
  • vSphere firewall / NSX microsegmentation — the attack is entirely internal to a single guest OS, no network path to block.
  • Patching only the ESXi host without updating VMware Tools in guests — VMware explicitly states the guest-side Tools update is required; the host patch alone is insufficient.
  • EDR on the hypervisor — there is no meaningful EDR for VMkernel and the bug isn't there anyway; detection has to sit in the Windows guest.
  • Snapshots / backups — do nothing to prevent EoP; they only help you recover after.
06 · Verification

Crowdsourced verification payload.

Run from a workstation with PowerCLI installed and a vCenter/ESXi credential that can enumerate hosts and VMs. Invoke as .\Check-CVE-2015-6933.ps1 -VIServer vcenter.corp.local. Requires read access to host build numbers and VM advanced settings; no changes are made.

noisgate-verify.ps1
POWERSHELLREAD-ONLYSAFE
#!powershell
# Check-CVE-2015-6933.ps1
# Detects ESXi 5.5 hosts below build 3248547 AND Windows guests with HGFS Shared Folders enabled.
# Exit codes: 0 = PATCHED, 1 = VULNERABLE, 2 = UNKNOWN

param(
    [Parameter(Mandatory=$true)][string]$VIServer
)

$ErrorActionPreference = 'Stop'
$fixedBuild = 3248547
$vulnerable = $false
$unknown    = $false

try {
    if (-not (Get-Module -ListAvailable VMware.PowerCLI)) {
        Write-Error 'VMware.PowerCLI module not installed.'
        exit 2
    }
    Import-Module VMware.PowerCLI -ErrorAction Stop | Out-Null
    Set-PowerCLIConfiguration -InvalidCertificateAction Ignore -Confirm:$false | Out-Null
    Connect-VIServer -Server $VIServer -ErrorAction Stop | Out-Null
} catch {
    Write-Host "UNKNOWN: could not connect to $VIServer -> $_"
    exit 2
}

$hosts = Get-VMHost | Where-Object { $_.Version -like '5.5.*' }
foreach ($h in $hosts) {
    $build = [int]($h.Build)
    if ($build -lt $fixedBuild) {
        Write-Host "VULNERABLE HOST: $($h.Name) build $build < $fixedBuild"
        $vulnerable = $true

        $vms = Get-VM -Location $h | Where-Object { $_.Guest.OSFullName -match 'Windows' }
        foreach ($vm in $vms) {
            $hgfs = Get-AdvancedSetting -Entity $vm -Name 'isolation.tools.hgfs.disable' -ErrorAction SilentlyContinue
            if (-not $hgfs -or $hgfs.Value -ne 'TRUE') {
                Write-Host "  -> Guest $($vm.Name) has HGFS NOT explicitly disabled"
            }
        }
    } else {
        Write-Host "PATCHED HOST: $($h.Name) build $build"
    }
}

Disconnect-VIServer -Confirm:$false | Out-Null

if ($vulnerable) { exit 1 }
elseif ($unknown) { exit 2 }
else { Write-Host 'PATCHED'; exit 0 }
07 · Bottom Line

If you remember one thing.

TL;DR
Do not lose sleep over this one. Under the noisgate mitigation SLA for LOW there is no mitigation deadline — go straight to the noisgate remediation SLA 365-day backlog window. Monday morning: run the PowerCLI check to inventory any ESXi 5.5 hosts still alive in your estate (there shouldn't be many — 5.5 has been out of Broadcom support since 2020-09-19). Fold every hit into your existing ESXi 5.5 decommission / migration project rather than patching in place. As a free-and-fast tactical control, push isolation.tools.hgfs.disable = "TRUE" to all Windows guests via PowerCLI in the next routine change window — that kills the attack surface for this CVE and a small family of related HGFS bugs at zero operational risk. Do not preempt Q3 patch cycles for MS17-010, PrintNightmare, or KEV-listed ESXi bugs to chase this one.

Sources

  1. Tenable plugin 87942
  2. NVD CVE-2015-6933
  3. VMware VMSA-2016-0001 (Broadcom)
  4. MITRE CVE-2015-6933
  5. Tenable VMSA-2016-0001 summary plugin
  6. CISA KEV catalog
  7. FIRST EPSS
  8. VMware ESXi 5.5 lifecycle (Broadcom)
Peer Review

What defenders are saying.

Submit a review attribution: handle + country only
0 flags selected · stored anonymously
Validation Results

Crowdsourced verification outputs.

Results submitted by users who ran the verification payload against their environment.