← Back to Feed CACHED · 2026-09-16 19:36:01 · CACHE_KEY CVE-2026-91843
CVE-2026-91843 · CWE-121 · Disclosed 2026-09-16

A stack overflow during the unauthenticated login process may

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

Someone left the master key under the doormat — and the door is made of cardboard

CVE-2026-91843 is a pre-authentication stack-based buffer overflow (CWE-121) in the login handler of Check Point's Security Management Server, Multi-Domain Security Management Server, Log Server, and Multi-Domain Log Server. An attacker sends an oversized username string to the management interface (typically TCP 18264/19009); the fwm process copies it onto the stack without bounds checking, overwriting the return address. No credentials, no user interaction, no special configuration — just a single crafted packet. Affected versions span R80 through R82.20 (with specific Jumbo Hotfix take thresholds), covering essentially every supported and recently-EOL release. Smart-1 Cloud is not affected.

The vendor's CVSS 9.8 CRITICAL rating is honest and, if anything, undersells the real-world consequence. This isn't a firewall data-plane bug that crashes a gateway — it's root-level code execution on the management plane, the box that holds every firewall policy, every admin credential, every VPN PSK, every certificate, and every audit log in your Check Point estate. Compromising the SMS is functionally equivalent to owning every gateway it manages. The vendor severity is accurate.

"Pre-auth root RCE on Check Point SMS. Your firewall fleet hangs in the balance."
02 · The Attack Path

5 steps from start to impact.

STEP 01

Reach the management interface

The attacker needs TCP connectivity to the SMS on port 18264 or 19009. In best-practice deployments, the SMS sits on a dedicated management VLAN behind ACLs. However, Check Point's implied rules historically allow SmartConsole connections from 'any,' and CheckMates community threads confirm many orgs have these ports reachable from broader networks or even the internet.
Conditions required:
  • Network path to SMS management port (TCP 18264 or 19009)
Where this breaks in practice:
  • Well-segmented environments restrict this to a jump host or admin VLAN
  • Not all deployments expose management ports to untrusted networks
Detection/coverage: Shodan/Censys queries for Check Point management ports; GreyNoise may flag scanning activity on 18264/19009.
STEP 02

Send the overflow payload

The attacker crafts a login request with an oversized username field. The fwm daemon's authentication handler uses an unbounded copy into a stack buffer. No authentication token or session is needed — the overflow occurs before any credential check. This is a classic CWE-121 stack smash. No weaponized tool is publicly available yet, but the primitive is textbook — skilled attackers can develop a working exploit from the advisory description alone.
Conditions required:
  • Ability to send a crafted TCP payload to the management port
Where this breaks in practice:
  • Stack canaries and ASLR on Gaia OS may complicate reliable exploitation
  • No public PoC exists as of 2026-09-17
Detection/coverage: SmartConsole audit log entry: 'Administrator failed to log in: Username too long.' IDS/IPS signatures for oversized SIC/SmartConsole auth payloads.
STEP 03

Achieve code execution as root

The fwm process runs as root on Gaia OS. Successful stack smash gives the attacker a root shell on the Security Management Server. No privilege escalation step is needed — fwm is already running at the highest privilege level.
Conditions required:
  • Working exploit that defeats stack protections on Gaia OS
Where this breaks in practice:
  • Requires bypass of ASLR/stack canaries (feasible but adds development time)
Detection/coverage: Process-level monitoring for anomalous child processes of fwm; unexpected outbound connections from the SMS.
STEP 04

Harvest the management plane

With root on the SMS, the attacker extracts all firewall policies, administrator password hashes, ICA certificates, SIC trust keys, VPN pre-shared keys, and the complete audit log. They can also push modified policies to every managed gateway. No additional tools needed — standard Check Point CLI utilities suffice.
Conditions required:
  • Root shell on SMS (achieved in Step 3)
Where this breaks in practice:
  • None — this is the normal operating capability of the SMS
Detection/coverage: Policy push audit logs; unexpected cpmi or fwm database exports; file integrity monitoring on $FWDIR/conf.
STEP 05

Pivot to managed gateways — fleet compromise

Using harvested SIC trust certificates and keys, the attacker establishes trusted management connections to every gateway in the domain, pushes rogue policies to open network paths, disables IPS blades, or installs backdoor access rules. This is fleet compromise — every firewall the SMS manages is now attacker-controlled.
Conditions required:
  • SIC trust material from Step 4
Where this breaks in practice:
  • None once SMS is owned — gateways trust their management server by design
Detection/coverage: Unexpected policy installations on gateways; SmartEvent alerts for policy changes outside change windows.
03 · Intelligence Metadata

The supporting signals.

In-the-wild exploitationNo confirmed exploitation as of 2026-09-17. Check Point has not reported active attacks.
Proof-of-conceptNo public PoC. Textbook CWE-121 stack overflow via oversized username — expect weaponization within days to weeks.
EPSS scoreNot yet scored (disclosed 2026-09-16, <24 hours old).
KEV statusNot listed as of 2026-09-17.
CVSS vectorCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H — worst-case pre-auth RCE profile: network-reachable, low complexity, no privileges, no interaction, full CIA impact.
Affected versionsR82.20; R82.10 (≤ Jumbo Take 44); R82 (≤ Take 126); R81.20 (≤ Take 166); R81.10 EOS (≤ Take 190); R81, R80.40, R80.30, R80.20, R80.10, R80 (all EOS).
Fixed versionsR82.20 Urgent Security Update Take 29; R82.10/R82/R81.20 Urgent Take 28. LivePatch available for in-memory patching. Smart-1 Cloud already patched.
Exposure dataCheck Point management ports (18264, 19009) visible on Shodan/Censys. Implied rules historically accept SmartConsole from 'any.' CheckMates community threads confirm internet exposure is common.
Disclosed2026-09-16 via Check Point advisory SK1000155.
Prior art / threat contextCVE-2026-16232 (SmartConsole auth bypass) was KEV-listed July 2026 — proves adversaries are actively hunting Check Point management plane bugs. Rapid7 released a public PoC for that flaw.
04 · The Call

noisgate verdict.

Final Verdict
= UNCHANGED to CRITICAL (9.8/10)

The affected component is the Check Point Security Management Server — the canonical trust anchor for every managed gateway, making fleet compromise the direct and immediate consequence of exploitation. The single most decisive factor is the role multiplier: pre-auth root RCE on the management plane means every firewall policy, admin credential, and SIC trust key is attacker-controlled, and modified policies can be pushed to every managed gateway without further exploitation.

HIGH Vulnerability mechanics and blast-radius assessment
MEDIUM Exploitation timeline — no PoC yet but trivial CWE-121 primitive

Why this verdict

  • Pre-auth, zero-interaction, network-reachable: CVSS:3.1/AV:N/AC:L/PR:N/UI:N — the attack surface is as broad as it gets. No credentials, no phishing, no user click. Baseline vendor score of 9.8 is warranted for the technical capability.
  • Root-level impact with no escalation needed: The fwm process runs as root on Gaia OS. A successful overflow is instant root. No second-stage privilege escalation step required.
  • Role multiplier — SMS is a fleet-control node by definition: The Security Management Server is not a workstation or a line-of-business app. 100% of SMS installations occupy the high-value role because that IS the role. Compromising it yields every firewall policy, admin credential, ICA cert, SIC trust key, and VPN PSK — and lets the attacker push modified policies to every managed gateway. Blast radius: fleet-scale. The verdict floor is CRITICAL per the role-multiplier rule.
  • Prior targeting of this exact attack surface: CVE-2026-16232 (SmartConsole auth bypass, Rapid7 PoC published) was KEV-listed in July 2026 with a 3-day CISA remediation deadline, proving adversaries are actively hunting Check Point management plane bugs. This makes exploitation of CVE-2026-91843 a matter of when, not if.
  • Friction audit — delays but does not reduce impact: Network segmentation (management VLAN) reduces reachability for internet-origin attackers, but community evidence and Shodan data confirm a significant fraction of deployments expose management ports. Even in segmented environments, any attacker with internal network access (post-phish, post-VPN, lateral movement) reaches the SMS. No public PoC exists yet (2026-09-17), but CWE-121 stack overflow via oversized input is one of the most well-understood exploit primitives — development time is days, not months.

Why not higher?

9.8 is already the CVSS ceiling for Scope:Unchanged. The CVSS spec does not go higher. If Scope:Changed were applicable (arguable given the gateway fleet impact), the score would exceed 10.0, which is capped. There is no room to upgrade.

Why not lower?

Every friction point (network segmentation, no public PoC yet, ASLR/stack canaries) is a delay, not a barrier. The SMS is a high-value-role component by definition — 100% of SMS installations occupy the high-value role because that IS the role. The blast radius (fleet compromise via policy push) means the verdict floor is CRITICAL per the role-multiplier rule. Downgrading past the floor would require evidence that <1% of the installed base occupies the high-value role, which is impossible for a product whose sole purpose is centralized firewall management. The July 2026 KEV listing of CVE-2026-16232 on the same attack surface confirms active adversary interest.

05 · Compensating Control

What to do — in priority order.

  1. Restrict SmartConsole Trusted Clients to approved IPs immediately — In SmartConsole → Manage & Settings → Permissions & Administrators → Trusted Clients, lock down to specific admin workstation IPs or a jump-host subnet. This removes network reachability for untrusted sources and is the single highest-impact compensating control. Deploy within the noisgate CRITICAL mitigation SLA of 3 days.
  2. Deploy Check Point LivePatch on all SMS and Log Servers — LivePatch patches the fwm process in memory without a reboot. Run cplp list in Expert mode — the patch should show 'armed' status with CVE-2026-91843 in the comment field. Deploy within 3 days.
  3. Block management ports at the perimeter — Ensure TCP 18264, 18190, 19009 are not reachable from the internet or untrusted network segments via perimeter firewall rules upstream of the Check Point estate. Deploy within 3 days.
  4. Monitor for exploitation attempts — Search SmartConsole audit logs for 'Administrator failed to log in: Username too long.' Set up alerting on this string. Any hit is a confirmed exploitation attempt — trigger incident response immediately.
  5. Migrate end-of-support releases — R80.x, R81, and R81.10 are EOS and will not receive patches. Plan migration to R81.20+ or R82.x within the noisgate CRITICAL remediation SLA of 90 days.
What doesn't work
  • IPS blade on the SMS itself — the SMS is the target, not a gateway inline with traffic. The IPS blade runs on Security Gateways, not on the management server's own login handler.
  • MFA on SmartConsole login — the overflow occurs *before* authentication, so MFA never fires. The buffer is smashed during username parsing, not after credential validation.
  • Gaia OS host firewall (iptables) — unless you've manually added custom iptables rules on the SMS (rare), the Gaia host firewall accepts management connections by implied rule. Trusted Client restrictions in SmartConsole are the correct layer.
06 · Verification

Crowdsourced verification payload.

Run on each Check Point Security Management Server or Log Server via SSH in Expert mode (expert command). No additional privileges beyond Expert mode are needed. Example: chmod +x check_cve_2026_91843.sh && ./check_cve_2026_91843.sh

noisgate-verify.sh
BASHREAD-ONLYSAFE
#!/bin/bash
# check_cve_2026_91843.sh - Verify CVE-2026-91843 patch status
# Run on Check Point SMS/Log Server in Expert mode
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN

set -euo pipefail

echo "=== CVE-2026-91843 Patch Verification ==="
echo "Target: $(hostname) | Date: $(date -u +%Y-%m-%dT%H:%M:%SZ)"
echo ""

# Check if this is a Check Point system
if [ ! -f /etc/cp-release ]; then
    echo "UNKNOWN - Not a Check Point system"
    exit 2
fi

echo "System: $(cat /etc/cp-release)"
echo ""

# Check LivePatch status
if command -v cplp &>/dev/null; then
    LIVEPATCH_OUTPUT=$(cplp list 2>&1 || true)
    echo "LivePatch status:"
    echo "$LIVEPATCH_OUTPUT"
    echo ""
    if echo "$LIVEPATCH_OUTPUT" | grep -qi "CVE-2026-91843" && echo "$LIVEPATCH_OUTPUT" | grep -qi "armed"; then
        echo "PATCHED - LivePatch for CVE-2026-91843 is armed"
        exit 0
    fi
else
    echo "WARNING: cplp command not found - LivePatch may not be installed"
fi

# Check installed hotfix version
VERSION=$(cat /etc/cp-release 2>/dev/null | grep -oP 'R\d+\.?\d*' | head -1 || echo "")
TAKE=$(cpinfo -y fw1 2>/dev/null | grep -i "take" | grep -oP '\d+' | tail -1 || echo "0")

echo "Detected version: $VERSION, Take: $TAKE"
echo ""

case "$VERSION" in
    R82.20)
        if [ "$TAKE" -ge 29 ]; then
            echo "PATCHED - $VERSION Take $TAKE >= Take 29"
            exit 0
        else
            echo "VULNERABLE - $VERSION Take $TAKE < Take 29"
            exit 1
        fi
        ;;
    R82.10)
        if [ "$TAKE" -ge 45 ]; then
            echo "PATCHED - $VERSION Take $TAKE >= Take 45"
            exit 0
        else
            echo "VULNERABLE - $VERSION Take $TAKE <= Take 44"
            exit 1
        fi
        ;;
    R82)
        if [ "$TAKE" -ge 127 ]; then
            echo "PATCHED - $VERSION Take $TAKE >= Take 127"
            exit 0
        else
            echo "VULNERABLE - $VERSION Take $TAKE <= Take 126"
            exit 1
        fi
        ;;
    R81.20)
        if [ "$TAKE" -ge 167 ]; then
            echo "PATCHED - $VERSION Take $TAKE >= Take 167"
            exit 0
        else
            echo "VULNERABLE - $VERSION Take $TAKE <= Take 166"
            exit 1
        fi
        ;;
    R81.10|R81|R80.40|R80.30|R80.20|R80.10|R80)
        echo "VULNERABLE - $VERSION is end-of-support. No patch available. Migrate immediately."
        exit 1
        ;;
    *)
        echo "UNKNOWN - Unrecognized version: $VERSION"
        exit 2
        ;;
esac
07 · Bottom Line

If you remember one thing.

TL;DR
This is a drop-everything vulnerability. CVE-2026-91843 gives an unauthenticated attacker root-level code execution on the box that controls every Check Point firewall in your environment. Under the noisgate CRITICAL mitigation SLA, you have 3 days to deploy compensating controls: restrict SmartConsole Trusted Clients to approved IPs, deploy Check Point LivePatch, and block management ports (18264/19009) from untrusted networks. Verify each SMS and Log Server with cplp list — you need 'armed' status with CVE-2026-91843 in the comment. Under the noisgate CRITICAL remediation SLA, apply the full Jumbo Hotfix (R82.20 Take 29 or R82.10/R82/R81.20 Take 28) within 90 days. If you are running R80.x, R81, or R81.10 (end-of-support), there is no patch — begin migration planning this week and apply Trusted Client restrictions and LivePatch (if available) as a bridge. Monitor SmartConsole audit logs for 'Username too long' entries — any hit means someone is probing your management plane and you should escalate to incident response immediately. The absence of a public PoC buys you days, not weeks. Move now.

Sources

  1. Check Point Advisory SK1000155
  2. CyberSecurity News — Check Point Root Access Flaw
  3. SecurityOnline — Check Point Login Flaw CVE-2026-91843
  4. Strix AI — CVE-2026-91843 Analysis
  5. OffSeq Threat Radar — CVE-2026-91843
  6. Cryptika — Check Point Root Access Vulnerability
  7. CheckMates — Management Server Internet Exposure
  8. BleepingComputer — CVE-2026-16232 SmartConsole Zero-Day
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.