← Back to Feed CACHED · 2026-07-01 06:28:24 · CACHE_KEY CVE-2026-14138
CVE-2026-14138 · CWE-451 · Disclosed 2026-06-30

Inappropriate implementation in WebAppInstalls in Google Chrome on Windows prior to 150

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

A phishing lure dressed up as a legitimate installable web app

CVE-2026-14138 is an *inappropriate implementation* in Chrome's WebAppInstalls component — the subsystem that handles installable PWAs and the install/launch prompts. Per the advisory, versions of Chrome on Windows prior to 150.0.7871.47 rendered install-related UI in a way that a crafted page could exploit to mislead a user (CWE-451: User Interface (UI) Misrepresentation of Critical Information). No memory corruption, no sandbox escape — the payoff is a spoofed prompt that could trick a user into installing or launching something they didn't intend.

Google labels this MEDIUM (4.2) and the CVSS vector AV:N/AC:H/PR:N/UI:R/S:U/C:N/I:L/A:L matches the reality: network reachable, but high attack complexity, user interaction required, and impact capped at *low integrity / low availability, zero confidentiality*. In a fleet of 10,000 hosts running Chrome auto-update, this is a backlog item, not a fire. The vendor rating is honest; if anything real-world friction pushes it lower.

"Chrome WebAppInstalls UI spoofing that needs a high-complexity chain plus a duped user — auto-update handles this."
02 · The Attack Path

4 steps from start to impact.

STEP 01

Attacker hosts a crafted page

The attacker stands up a web property that abuses the WebAppInstalls flow — typically a page that invokes beforeinstallprompt handling or a manifest with attacker-chosen name/icons/start_url so the resulting Chrome install UI misrepresents origin or intent. Weaponization is essentially a static HTML+manifest payload; no CVE-specific exploit kit is required, just knowledge of the flawed UI path.
Conditions required:
  • Attacker-controlled domain reachable from the victim
  • Understanding of the specific WebAppInstalls UI misrender path
Where this breaks in practice:
  • AC:H — the misrepresentation only fires under specific manifest/timing conditions
  • Modern secure web gateways categorize newly-registered domains and block them
Detection/coverage: Zero scanner coverage for the page itself; URL reputation feeds (Cisco Umbrella, Zscaler, Netskope) will catch typical phishing infrastructure.
STEP 02

Lure the user to the page

Delivery is standard phishing plumbing — email link, malvertising, SEO poisoning, or a compromised third-party site. The user must actually *navigate* to the page in Chrome on Windows; this is not a drive-by against browsing metadata.
Conditions required:
  • A working delivery channel (email, ad, redirect)
  • Chrome on Windows as the victim browser
Where this breaks in practice:
  • Email security (Proofpoint, Mimecast, M365 Defender) strips or sandboxes the link
  • SmartScreen and Google Safe Browsing block known-bad reputations
Detection/coverage: Email gateway URL rewriting and click-time analysis; EDR web-content telemetry.
STEP 03

User interacts with the spoofed install prompt

The victim sees the misrepresented WebAppInstalls UI and clicks *Install* (or equivalent) believing it is a trusted app. The attacker gains a low-integrity foothold: an installed PWA that can display attacker content in an app-frame chrome, or influence a subsequent launch. No code execution outside Chrome's normal PWA sandbox.
Conditions required:
  • Victim clicks through the prompt
  • No enterprise policy blocking PWA install
Where this breaks in practice:
  • WebAppInstallForceList / DefaultWebAppSettings enterprise policies can disable installs entirely
  • Trained users increasingly ignore install prompts from unknown sites
Detection/coverage: Chrome enterprise reporting (Chrome Browser Cloud Management) logs PWA installs.
STEP 04

Post-install abuse

With an installed PWA the attacker can present a persistent app icon that launches attacker-controlled content in a windowed frame, feeding follow-on phishing (credential capture, MFA prompt bombing pages). This is where integrity:low / availability:low shows up — the attacker didn't read your files, they nudged your user.
Conditions required:
  • Victim launches the installed PWA later
  • Follow-on phishing infrastructure ready
Where this breaks in practice:
  • IdP MFA (FIDO2/WebAuthn) neutralizes credential replay
  • Conditional access blocks unknown-device sign-ins
Detection/coverage: IdP anomalous sign-in signals; EDR process-tree from chrome.exe --app-id= launches.
03 · Intelligence Metadata

The supporting signals.

In-the-wild statusNo known exploitation. Not listed in CISA KEV as of 2026-07-01.
Public PoCNone published on GitHub or exploit-db at time of assessment; Chromium bug tracker entry likely still restricted under Google's standard 14-week disclosure embargo.
EPSSExpected <0.5% — CWE-451 UI-spoofing Chrome bugs historically score in the bottom decile of EPSS.
KEV statusNot listed.
CVSS v3.1 vectorAV:N/AC:H/PR:N/UI:R/S:U/C:N/I:L/A:L → network, high complexity, user interaction required, no confidentiality impact.
Affected versionsGoogle Chrome on Windows prior to 150.0.7871.47. Chromium-derived browsers (Edge, Brave, Opera, Vivaldi) inherit until they merge the upstream fix.
Fixed versionChrome 150.0.7871.47 (Stable channel, released 2026-06-30).
Exposure telemetryChrome auto-update reaches ~95% of managed installs within 7 days per Google's own release telemetry — the exposed population decays fast without operator action.
Disclosure date2026-06-30 via Chrome Stable channel update advisory.
ReporterCredited external researcher per Google's advisory (name typically withheld pre-embargo).
04 · The Call

noisgate verdict.

Final Verdict
DOWNGRADED to LOW (3.5/10)

The single decisive factor is impact ceiling: the CVSS vector caps damage at low integrity / low availability with zero confidentiality impact, and the affected component is a user-space browser UI element — not an identity, hypervisor, or kernel-mode role. Chrome's silent auto-update further collapses the exposed population within days, so this belongs in the backlog-hygiene tier.

HIGH impact assessment (UI spoofing, no RCE)
HIGH exposed-population decay via Chrome auto-update
MEDIUM PoC availability (Chromium bug tracker still embargoed)

Why this verdict

  • Impact ceiling is trivialC:N/I:L/A:L. No code execution, no data theft, no privilege escalation. Worst case is a user tricked into installing a fake PWA that facilitates *follow-on* phishing already covered by other controls.
  • High attack complexityAC:H combined with UI:R means the attacker needs both a specific technical condition and a duped user. Two compounding friction points.
  • Role multiplier: workstation only. Chrome runs on user endpoints; there is no domain-controller, hypervisor, IdP, PAM, or backup-admin variant of this bug. The blast radius is capped at the single user's browser session — no fleet, tenant, or domain pivot. Floor stays at LOW.
  • Auto-update is the world's largest compensating control for Chrome. Enterprises on Chrome Browser Cloud Management or default Windows Update-for-Business channels will be patched before most teams finish reading the advisory.
  • Compensating tooling is dense — Safe Browsing, SmartScreen, secure web gateway URL filtering, email link rewriting, MFA/FIDO2, and existing user phishing training all sit in the attack path.

Why not higher?

MEDIUM would be defensible if the impact vector included confidentiality or if there were an unpatched-population signal. Neither exists: the vector is C:N and auto-update rapidly drains exposure. Nothing in the chain terminates in a fleet-scale outcome.

Why not lower?

IGNORE is wrong because the bug is real, present in an internet-exposed attack surface (the browser), and phishing chains that end in credential theft do use trust-shifting UI tricks like this. It deserves the standard backlog patch, not a rationale-only dismissal.

05 · Compensating Control

What to do — in priority order.

  1. Let Chrome auto-update do its job — Confirm ChromeAutoUpdate and UpdateDefault GPO/registry settings are enabled fleet-wide; verify Chrome Browser Cloud Management reporting shows >95% of hosts on 150.0.7871.47 or later within 14 days. LOW verdict has no mitigation SLA — you have the full 365-day remediation window, but auto-update should close this in a week regardless.
  2. Restrict PWA installation via enterprise policy — Set DefaultWebAppSettings to block_install or use WebAppInstallForceList to allow-list only approved PWAs. Deployed via GPO/MDM this neutralizes the exploit's payoff even on unpatched hosts.
  3. Enforce Safe Browsing Enhanced Protection — Push SafeBrowsingProtectionLevel=2 via Chrome policy. Google's Safe Browsing catches the phishing lure well before the WebAppInstalls UI ever renders.
  4. Confirm FIDO2/WebAuthn on identity provider — Since the terminal harm is credential-phishing amplification, phishing-resistant MFA (Okta FastPass, Entra passkeys, YubiKey) means a spoofed PWA cannot cash out into an account takeover.
What doesn't work
  • EDR alone — process telemetry does not detect a legitimate chrome.exe rendering a misleading install prompt. There's no malicious binary to flag.
  • Network IDS signatures — the payload is HTML + a web manifest over TLS; nothing on the wire distinguishes benign PWA install traffic from the exploit.
  • Blocking Chrome entirely — Chromium-derived browsers (Edge, Brave, Opera) inherit the flaw until upstream merges land. Swapping browsers doesn't fix it.
06 · Verification

Crowdsourced verification payload.

Run on any Windows host as a standard user (elevation not required). Example: .\Check-CVE-2026-14138.ps1. Detects installed Chrome version across per-user and per-machine install paths and compares to the fixed build 150.0.7871.47.

noisgate-verify.ps1
POWERSHELLREAD-ONLYSAFE
#Requires -Version 5.0
# Check-CVE-2026-14138.ps1
# Detects Google Chrome versions vulnerable to CVE-2026-14138 (WebAppInstalls UI spoof)
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN/NotInstalled

$fixed = [Version]'150.0.7871.47'
$paths = @(
    "$env:ProgramFiles\Google\Chrome\Application\chrome.exe",
    "${env:ProgramFiles(x86)}\Google\Chrome\Application\chrome.exe",
    "$env:LOCALAPPDATA\Google\Chrome\Application\chrome.exe"
)

$found = $false
$vuln  = $false

foreach ($p in $paths) {
    if (Test-Path $p) {
        $found = $true
        try {
            $v = [Version](Get-Item $p).VersionInfo.ProductVersion
            Write-Host "[+] Chrome found at $p -> $v"
            if ($v -lt $fixed) {
                Write-Host "    VULNERABLE (fixed in $fixed)" -ForegroundColor Red
                $vuln = $true
            } else {
                Write-Host "    PATCHED" -ForegroundColor Green
            }
        } catch {
            Write-Host "[!] Could not parse version at $p" -ForegroundColor Yellow
        }
    }
}

if (-not $found) {
    Write-Host 'UNKNOWN: Google Chrome not installed on this host'
    exit 2
}
if ($vuln) { Write-Host 'VULNERABLE'; exit 1 }
Write-Host 'PATCHED'; exit 0
07 · Bottom Line

If you remember one thing.

TL;DR
Monday morning: verify Chrome auto-update is enabled fleet-wide and confirm your Chrome Browser Cloud Management dashboard shows the population converging on 150.0.7871.47. This is a LOW verdict — per the noisgate mitigation SLA there is no mitigation deadline (go straight to the noisgate remediation SLA window of ≤365 days), but auto-update will close this within 7-14 days without operator action. Spend the saved cycles pushing SafeBrowsingProtectionLevel=2 and locking down DefaultWebAppSettings via policy — those controls defang an entire class of PWA-based phishing, not just this one CVE.

Sources

  1. Chrome Releases – Stable Channel Update for Desktop
  2. NVD entry for CVE-2026-14138
  3. CWE-451: User Interface (UI) Misrepresentation of Critical Information
  4. Chromium Issue Tracker (embargoed bugs list)
  5. Chrome Enterprise policy: DefaultWebAppSettings
  6. Chrome Enterprise policy: SafeBrowsingProtectionLevel
  7. CISA Known Exploited Vulnerabilities Catalog
  8. FIRST EPSS scoring model
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.