This is a second-story window problem, not a front-door break-in
CVE-2026-0107 is a local elevation-of-privilege bug in the Pixel-side CPM component, described as a confused-deputy flaw in gmc_ddr_handle_mba_mr_req within gmc_mba_ddr.c. Google lists it in the Pixel Update Bulletin—March 2026 as a High severity EoP issue, and devices with a 2026-03-05 or later security patch level are stated to address the bulletin issues. The affected range is therefore best understood as supported Google Pixel devices missing the March 5, 2026 patch level, not the whole internet and not every Android build.
Vendor HIGH 8.4 is technically defensible in a lab because successful exploitation could hand over high confidentiality, integrity, and availability impact on the device. In enterprise reality, though, the decisive friction is attacker position: this is AV:L and effectively requires code running on the handset first. That makes it a post-initial-access privilege escalator for a subset of Android fleets, so it deserves attention but not the same urgency as remotely reachable mobile RCEs or browser/BBU/baseband bugs with exploitation evidence.
4 steps from start to impact.
Land code on the handset
adb-assisted deployment rather than an internet exploit kit.- Attacker can execute code locally on the Android/Pixel device
- Target device is missing the 2026-03-05 security patch level
- Target is in the affected Pixel population using the vulnerable CPM path
- Requires prior foothold on-device, which is already a major stage in the intrusion chain
- Managed Play, app allowlisting, MDM enrollment, and Play Protect reduce delivery options
- Not every Android device is affected; Google published it in the Pixel bulletin under
CPM
Trigger the confused deputy in CPM
gmc_ddr_handle_mba_mr_req so a more-privileged component performs work on the attacker's behalf. In practice this is likely a purpose-built Binder/ioctl/native harness, not a commodity off-the-shelf module. No public Google patch reference or public PoC was surfaced in the advisory material I found.- The vulnerable request path is reachable from the attacker's local context
- The attacker can craft the expected request structure to the privileged intermediary
- Public exploit details appear sparse because Google marks the referenced bug as non-public
- Driver/firmware-specific behavior often makes reliable exploitation device-build dependent
adb shell misuse, unusual native process activity, or suspicious driver access. Signature coverage is likely weak until public PoC details emerge.Elevate privileges on-device
- Exploit chain successfully coerces the privileged component
- Target build lacks the March 2026 fix
- Modern Android hardening, SELinux policy, and vendor-specific implementation details can reduce exploit reliability
- The blast radius is usually the single device already under attack, not an enterprise-wide service
Use the new privileges for post-compromise objectives
- Attacker already controls a local process on the device
- Device contains enterprise apps, tokens, messages, or VPN material worth stealing
- Compromise remains centered on the already-breached handset
- Conditional access, strong app attestation, and short-lived tokens can limit downstream enterprise impact
The supporting signals.
| In-the-wild status | No public evidence of active exploitation was found in the vendor bulletin material or CISA KEV. Treat as not known exploited as of the March 2026 disclosure set. |
|---|---|
| KEV status | Not KEV-listed. CISA's catalog does not indicate inclusion for this CVE. |
| Proof-of-concept availability | No public PoC was surfaced in the vendor references or common CVE aggregators reviewed. That does not prove exploit absence, but it does reduce copy-paste attacker pressure. |
| EPSS | 0.00008 from the user-supplied intel, which is effectively near-floor exploit likelihood in the next 30 days. |
| CVSS vector reality check | CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H means no privileges and no user interaction once local code exists, but the key limiter is still AV:L: the attacker must already be on the device. |
| Affected scope | Google lists this in the Pixel Update Bulletin—March 2026 as EoP / High / CPM, implying a Google-device-specific exposure set rather than all Android endpoints. |
| Fixed version / patch level | For Google devices, security patch level 2026-03-05 or later addresses bulletin issues, including this one. |
| Affected version range | Best current operational range: supported Pixel devices below SPL 2026-03-05. Google does not publish a neat semantic version for the vulnerable CPM component in the bulletin. |
| Scanning / exposure data | No meaningful Shodan/Censys/GreyNoise lens here because this is not an internet-facing service. Exposure must be measured from EMM/MDM inventory and handset patch posture, not external scanning. |
| Disclosure and source | Disclosed 2026-03-10 by Google_Devices in the CVE record; Pixel bulletin published 2026-03-03 and updated 2026-03-09. |
noisgate verdict.
The single biggest reason this lands in MEDIUM is that exploitation is post-foothold: the attacker already needs local code execution on the handset. That sharply narrows both the reachable population and the enterprise blast radius compared with remotely exploitable Android flaws, even though the on-device impact can be severe.
Why this verdict
- Start at vendor 8.4, then cut for attacker position:
AV:Lmeans this is not an initial-access bug; the attacker must already have code on the device. - Population narrows again: Google lists the issue in the Pixel bulletin under
CPM, so this is not a generic all-Android internet problem. - No exploitation pressure: no KEV listing, no public exploitation evidence surfaced, and the supplied EPSS 0.00008 is extremely low.
Why not higher?
It is not higher because this is not remotely reachable and it does not create a fleet-wide worm path. Every real attack chain must first solve app delivery, sideloading, physical access, or some other local foothold problem before this CVE matters.
Why not lower?
It is not lower because once the attacker is on the device, the vulnerability can materially increase control and defeat the normal app sandbox boundary. For enterprises with managed Pixel fleets, local EoP bugs still matter because they turn a modest handset foothold into access to enterprise tokens, messages, VPN clients, and sensitive mobile data.
What to do — in priority order.
- Query mobile patch posture — Use your EMM/MDM to identify Pixel devices below
2026-03-05SPL and map them to corporate-owned, BYOD, and high-risk user groups. Because this is a MEDIUM verdict, there is no mitigation SLA — go straight to the 365-day remediation window, but you should still inventory exposure immediately. - Restrict app sources — Enforce managed Google Play, block unknown sources where policy allows, and tighten app allowlisting for privileged or regulated user populations. This directly attacks the main prerequisite — getting local code onto the handset — and should be maintained while devices move through normal update cadence.
- Disable developer-risk settings — Use mobile policy to limit
adb, developer options, and USB debugging except for approved engineering populations. That removes one of the easiest operator-assisted paths to local exploitation and should stay in place through the remediation window. - Use device trust in conditional access — Gate access to sensitive SaaS, VPN, and admin apps on mobile compliance signals such as patch level, Play Protect state, and management status. This reduces downstream blast radius if a handset is locally compromised before remediation is complete.
- Prioritize risky cohorts first — Patch executive, admin, developer, travel, and high-value target groups first inside your normal mobile OS rollout. Even without a mitigation SLA for MEDIUM, those users carry disproportionate credential and data value during the 365-day remediation window.
- Perimeter firewalls or internet IDS: this is AV:L, so network edge controls do not stop exploitation on the device.
- Email filtering alone: it may reduce one delivery path for a malicious app or link, but it does nothing once local code is already present.
- Waiting for vulnerability scans from traditional VM tools: most enterprise scanners do not see handset patch posture well enough; you need mobile management telemetry.
Crowdsourced verification payload.
Run this from an auditor workstation with adb installed and USB or managed debugging access to the target Android/Pixel device. Invoke it as ./check_cve_2026_0107.sh SERIAL123 or omit the serial if only one device is connected; it needs no root, but it does need permission to use adb shell getprop against the handset.
#!/usr/bin/env bash
# check_cve_2026_0107.sh
# Verify likely exposure to CVE-2026-0107 on Android/Pixel devices.
# Logic: Google states Pixel devices with security patch level 2026-03-05 or later
# address the March 2026 Pixel bulletin issues. This script checks that SPL.
# Output: VULNERABLE / PATCHED / UNKNOWN
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN, 3=usage/dependency error
set -euo pipefail
REQ_SPL="2026-03-05"
SERIAL="${1:-}"
fail() {
echo "UNKNOWN: $1"
exit 2
}
need_cmd() {
command -v "$1" >/dev/null 2>&1 || {
echo "UNKNOWN: missing dependency '$1'" >&2
exit 3
}
}
need_cmd adb
need_cmd awk
need_cmd sed
ADB=(adb)
if [[ -n "$SERIAL" ]]; then
ADB+=( -s "$SERIAL" )
fi
# Ensure device is reachable
STATE="$(${ADB[@]} get-state 2>/dev/null || true)"
if [[ "$STATE" != "device" ]]; then
fail "adb target not reachable or not in 'device' state"
fi
getprop_safe() {
local key="$1"
${ADB[@]} shell getprop "$key" 2>/dev/null | tr -d '\r'
}
MANUFACTURER="$(getprop_safe ro.product.manufacturer | sed 's/^ *//;s/ *$//')"
BRAND="$(getprop_safe ro.product.brand | sed 's/^ *//;s/ *$//')"
MODEL="$(getprop_safe ro.product.model | sed 's/^ *//;s/ *$//')"
SPL="$(getprop_safe ro.build.version.security_patch | sed 's/^ *//;s/ *$//')"
FINGERPRINT="$(getprop_safe ro.build.fingerprint | sed 's/^ *//;s/ *$//')"
if [[ -z "$SPL" ]]; then
fail "security patch level not available from device properties"
fi
# Validate date format YYYY-MM-DD
if ! [[ "$SPL" =~ ^[0-9]{4}-[0-9]{2}-[0-9]{2}$ ]]; then
fail "unexpected SPL format: '$SPL'"
fi
# Compare ISO dates lexicographically; safe for YYYY-MM-DD
if [[ "$SPL" < "$REQ_SPL" ]]; then
echo "VULNERABLE: model='$MODEL' manufacturer='$MANUFACTURER' brand='$BRAND' spl='$SPL' required>='$REQ_SPL'"
echo "DETAIL: fingerprint='$FINGERPRINT'"
exit 1
else
echo "PATCHED: model='$MODEL' manufacturer='$MANUFACTURER' brand='$BRAND' spl='$SPL' required>='$REQ_SPL'"
echo "DETAIL: fingerprint='$FINGERPRINT'"
exit 0
fi
If you remember one thing.
2026-03-05 SPL, then patch them through the normal mobile OS process while tightening app-source and developer-mode policy for higher-risk users. Because this is MEDIUM and there is noisgate mitigation SLA for that bucket, no mitigation SLA — go straight to the 365-day remediation window; the noisgate remediation SLA is ≤365 days, so this belongs in your routine mobile patch program rather than your emergency change queue.Sources
- Google Pixel Update Bulletin — March 2026
- Android Security Bulletin — March 2026
- CISA Known Exploited Vulnerabilities Catalog
- CWE-441: Unintended Proxy or Intermediary ('Confused Deputy')
- Google Help: Check & update your Android version
- CVE Record — CVE-2026-0107
- OpenCVE mirror for CVE-2026-0107 metadata
- Google Android Enterprise device trust signals
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.