Someone found the master key to your firewall control room and the door was never locked
CVE-2026-18574 is an authentication bypass (CWE-288) in Check Point Security Management Server and Multi-Domain Security Management Server (MDS) that allows an unauthenticated remote attacker with network access to management services to execute arbitrary commands. The flaw spans a massive version range: R80, R80.10, R80.20, R80.30, R80.40, R81, R81.10, R81.20, R82, and R82.10. Patches exist only for R81.20+, meaning the R80.x and early R81.x branches (all end-of-support) have no fix. Check Point discovered it internally and published it on 2026-08-03 with a CVSS 4.0 score of 9.3 CRITICAL.
The vendor severity of 9.3 is honest and, if anything, could be *understated* for shops that haven't segmented their SMS. This is not a theoretical policy-read or info-leak — it's full command execution on the box that authors every firewall rule, holds every gateway SIC certificate, and stores administrator credentials. The only friction is network reachability to ports 18190/19009, which is restricted in well-segmented environments but trivially reachable from any host on the management VLAN or through misconfigured implied rules. The blast radius — rewriting firewall policy across every managed gateway — is fleet-scale by definition.
4 steps from start to impact.
Reach Management Services
Any to management services, and Shodan/Censys regularly index exposed instances. An internal attacker post-initial-access on any host that can route to the SMS qualifies.- Network path to SMS on port 18190/19009/18264
- Management VLAN segmentation blocks lateral traffic from user/server VLANs
- Best-practice Trusted Clients config restricts source IPs
- Internet-facing SMS is rare (but not zero — CheckMates forum threads document it)
Exploit Authentication Bypass
- Vulnerable SMS version (R80–R82.10 without hotfix)
- No public PoC exists as of 2026-08-07, raising the attacker skill bar temporarily
- Check Point IPS signatures may detect exploitation attempts if the SMS is behind a managed gateway
cpd/fwm process activity or unexpected SmartConsole login events.Execute Arbitrary Commands
$FWDIR/conf, the management database, SIC certificates, and administrator password hashes.- Successful auth bypass from step 2
- None — once auth is bypassed, command execution is the documented impact
/var/log/messages, cpwd_admin), unexpected cpconfig or mgmt_cli invocations, file integrity monitoring on $FWDIR/conf/objects_5_0.C and policy files.Modify Firewall Policies or Pivot
- Command execution on SMS (step 3)
- Policy installation generates SmartConsole audit trail and gateway logs
- Change-management processes may detect unexpected policy pushes — but only after the fact
fw log showing unexpected policy timestamps. Syslog forwarding of SMS audit events to external SIEM.The supporting signals.
| In-the-Wild Exploitation | No evidence. Check Point states the vulnerability was discovered internally with no observed exploitation as of 2026-08-03. |
|---|---|
| Proof-of-Concept | No public PoC found on GitHub, Exploit-DB, or researcher blogs as of 2026-08-07. The related CVE-2026-16232 (SmartConsole auth bypass) has a Rapid7 PoC, raising the likelihood that researchers will produce one for this CVE soon. |
| EPSS | 0.00991 (≈1.0%) — low current exploitation probability, but this predates wide awareness of the CVE. |
| KEV Status | Not listed on CISA KEV as of 2026-08-07. |
| CVSS Vector | CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N — 9.3 CRITICAL. Network-accessible, no auth, no complexity, full CIA impact. |
| Affected Versions | R80, R80.10, R80.20, R80.30, R80.40, R81, R81.10, R81.20, R82, R82.10 — essentially every Gaia release in the last decade. |
| Fixed Versions | R82.10 Take 404040+, R82 Take 122122122+, R81.20 Take 161161161+. No patches for R80.x or R81/R81.10 (end-of-support). Smart-1 Cloud already protected. |
| Exposure Data | Check Point management ports (18190, 19009, 18264) are regularly indexed by Shodan/Censys. CheckMates community threads confirm implied rules can expose these ports to Any. Exact internet-facing count not published, but the primary risk is internal network reachability post-initial-access. |
| Disclosure Date | 2026-08-03 (Check Point advisory sk185222) |
| Reporting Researcher | Not disclosed — Check Point states the vulnerability was discovered internally. |
noisgate verdict.
The single most decisive factor is role identity: Check Point Security Management Server *is* the central trust anchor for every managed firewall gateway — compromise equals fleet-wide policy control, SIC certificate theft, and VPN credential exfiltration. An unauthenticated, low-complexity RCE on a component that is *definitionally* a high-value-role asset (network security management plane) meets the CRITICAL floor with no friction sufficient to override it.
Why this verdict
- Unauthenticated RCE, zero complexity: CVSS:4.0 AV:N/AC:L/AT:N/PR:N/UI:N — the lowest possible attacker effort once network access is achieved. No credential theft, no phishing, no user interaction required.
- Role multiplier — SMS is the fleet's firewall brain: Check Point SMS is not a workstation or app server. It is the *sole* policy authority for every managed gateway. 100% of SMS installations occupy the high-value network-management-plane role by definition. Compromise yields: policy rewrite across all gateways, SIC certificate extraction, admin credential theft, VPN PSK exfiltration. Blast radius: fleet-scale.
- Enormous version surface with EoS gaps: Every Gaia release from R80 through R82.10 is affected. R80.x and R81/R81.10 are end-of-support with no patch available, forcing a major upgrade or compensating controls. Organizations running EoS versions cannot simply apply a hotfix.
- Adjacent CVE-2026-16232 PoC lowers the weaponization bar: A public PoC already exists for the related SmartConsole auth bypass (CVE-2026-16232, Rapid7). Researchers familiar with Check Point's SIC/CPMI internals are likely to produce a CVE-2026-18574 PoC rapidly.
Why not higher?
A 9.3 is already near-ceiling. The only reason this is not a 10.0 is that the CVSS 4.0 vector shows SC:N/SI:N/SA:N (no subsequent-system impact beyond the SMS itself in the formal model), and the vendor reports no active exploitation. In practice, the SMS *does* control subsequent systems (gateways), but the formal scoring reflects direct impact only.
Why not lower?
Despite management VLAN segmentation providing friction, the SMS is *canonically* a high-value-role component — 100% of installations are the firewall management plane. The unauthenticated, zero-complexity attack chain means any internal attacker (or any host compromised via phishing, VPN, or another CVE) that can route to the SMS can exploit this. Network segmentation is a compensating control, not a design guarantee — implied rules regularly break it. The CRITICAL floor applies and cannot be overridden by generic 'requires internal access' friction alone.
What to do — in priority order.
- Lock down Trusted Clients to explicit admin IPs immediately — In SmartConsole → Global Properties → Management Access → Trusted Clients, restrict GUI client access to a named list of administrator workstation IPs. This is the single highest-ROI control — it eliminates the network-reachability prerequisite for most internal attackers. Deploy within 3 days per noisgate mitigation SLA for CRITICAL.
- Block management ports at the network layer — Add explicit ACLs on switches/firewalls upstream of the SMS to permit TCP 18190, 19009, and 18264 only from the management jump host subnet. Do not rely on Check Point implied rules alone. Deploy within 3 days.
- Isolate SMS on a dedicated management VLAN with MFA jump host — If not already segmented, place the SMS on a VLAN reachable only through an MFA-gated jump host. This adds an authentication barrier the CVE cannot bypass. Deploy within 3 days for CRITICAL.
- Enable and forward SMS audit logs to external SIEM — Forward
/var/log/messages, SmartConsole audit logs, andcpddaemon logs to an external SIEM. Create alerts for: unexpected SmartConsole logins, policy installations outside change windows, andmgmt_clicommands from non-admin source IPs. This provides detection-in-depth. - Upgrade EoS versions (R80.x, R81, R81.10) immediately — These versions have no patch. The only remediation is upgrading to R81.20+ and applying the Jumbo Hotfix. Treat this as an emergency migration — begin planning within 3 days, execute within 90 days per noisgate remediation SLA.
- Check Point IPS blade on the SMS itself — the IPS blade runs on Security Gateways, not on the SMS. Even if a gateway sits in front of the SMS, the management traffic is typically permitted by implied rules and may not be inspected by IPS.
- MFA on SmartConsole login — this CVE bypasses the authentication mechanism entirely; MFA layered on top of SmartConsole auth is irrelevant because the attacker never goes through the normal login flow.
- Certificate-based SIC trust alone — CWE-288 specifically describes bypass via an alternate path or channel, meaning the normal SIC certificate validation path may be circumvented entirely.
Crowdsourced verification payload.
Run this script on the Check Point SMS/MDS appliance itself as the admin user in Expert mode (expert shell). Invoke with: bash check_cve_2026_18574.sh. Requires Expert mode access to read cpinfo and installed hotfix data.
#!/bin/bash
# check_cve_2026_18574.sh
# Checks if Check Point SMS/MDS is patched for CVE-2026-18574
# Run on the SMS/MDS appliance in Expert mode
# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN
set -euo pipefail
echo "=== CVE-2026-18574 Check ==="
echo "Date: $(date)"
echo "Hostname: $(hostname)"
# Get Gaia version
if command -v clish &>/dev/null; then
VERSION=$(clish -c 'show version product' 2>/dev/null | grep -oP 'R[0-9]+\.?[0-9]*' | head -1)
else
VERSION=$(cat /etc/cp-release 2>/dev/null | grep -oP 'R[0-9]+\.?[0-9]*' | head -1)
fi
if [ -z "$VERSION" ]; then
echo "UNKNOWN - Could not determine Gaia version."
exit 2
fi
echo "Detected version: $VERSION"
# Get installed Jumbo Take
JUMBO_TAKE=$(cpinfo -y fw1 2>/dev/null | grep -i 'jumbo' | grep -oP 'Take [0-9]+' | grep -oP '[0-9]+' | tail -1)
if [ -z "$JUMBO_TAKE" ]; then
JUMBO_TAKE=0
fi
echo "Installed Jumbo Take: $JUMBO_TAKE"
case "$VERSION" in
R82.10)
REQUIRED=404040
;;
R82)
REQUIRED=122122122
;;
R81.20)
REQUIRED=161161161
;;
R81.10|R81|R80.40|R80.30|R80.20|R80.10|R80)
echo "VULNERABLE - $VERSION is end-of-support with NO available patch for CVE-2026-18574."
echo "Action: Upgrade to R81.20+ and apply the required Jumbo Hotfix."
exit 1
;;
*)
echo "UNKNOWN - Unrecognized version: $VERSION"
exit 2
;;
esac
if [ "$JUMBO_TAKE" -ge "$REQUIRED" ]; then
echo "PATCHED - $VERSION Take $JUMBO_TAKE meets or exceeds required Take $REQUIRED."
exit 0
else
echo "VULNERABLE - $VERSION Take $JUMBO_TAKE is below required Take $REQUIRED."
echo "Action: Apply Jumbo Hotfix Accumulator Take $REQUIRED or later."
exit 1
fiIf you remember one thing.
Sources
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.