← Back to Feed CACHED · 2026-09-23 05:14:31 · CACHE_KEY CVE-2026-93616
CVE-2026-93616 · CWE-22 · Disclosed 2026-09-22

A directory traversal and file upload vulnerability

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

Handing an unauthenticated stranger the master console for your entire firewall fleet

CVE-2026-93616 is a pre-authentication path traversal and arbitrary file upload vulnerability in the Check Point Management web service (TCP/19009). An unauthenticated attacker who can reach this port can combine directory traversal with file upload handling to plant a malicious script or Java class in an arbitrary path, then trigger the management service to execute it. This yields code execution on the Security Management Server (SMS), Multi-Domain Security Management Server (MDS), Log Server, Multi-Domain Log Server, or SmartEvent — any product running the management web service. Affected versions span the entire supported product line: R82.20, R82.10 through JHF Take 44, R82 through Take 126, R81.20 through Take 166, R81.10 through Take 190, and all End-of-Support R80.x and R81 releases.

The vendor's CRITICAL / 9.8 rating is accurate and if anything understates the operational blast radius. The CVSS vector (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) correctly captures the pre-auth, zero-click, low-complexity nature of the flaw. What CVSS cannot express is that the affected component is *the central nervous system of your entire Check Point deployment* — owning the SMS means silently rewriting every firewall policy, harvesting every VPN credential, and pivoting to every managed gateway. Check Point confirmed active exploitation on July 23, 2026, two months before the public advisory, and CISA added the CVE to the KEV catalog on September 22, 2026 with a 3-day remediation deadline under BOD 26-04. This is a drop-everything-and-patch situation.

"Pre-auth RCE on your firewall management plane -- actively exploited, KEV-listed. Patch within hours."
02 · The Attack Path

5 steps from start to impact.

STEP 01

Network reachability to TCP/19009

The attacker identifies a Check Point Management Server with TCP port 19009 (Management web service / SmartConsole) reachable from their vantage point. This port is used for SmartConsole connections and CPM/DLE SOAP operations. In best-practice deployments it is firewalled to trusted admin IPs only, but implied rules and misconfigured 'Trusted Clients: Any' settings frequently leave it accessible from the internal network or even the internet.
Conditions required:
  • Network path to target SMS/MDS on TCP/19009
Where this breaks in practice:
  • Best-practice deployments restrict this port to named admin IPs
  • Internet-facing exposure requires a misconfiguration, though internal-network access is common
Detection/coverage: Firewall connection logs showing unexpected sources hitting TCP/19009. Shodan/Censys scans can identify externally exposed instances.
STEP 02

Path traversal via management web service

The attacker sends a crafted unauthenticated HTTP request to the management web service exploiting improper pathname validation (CWE-22). The request uses directory traversal sequences to escape the intended upload directory. No credentials, tokens, or session cookies are required. Indicators include usernames exceeding 1,000 characters in cpm.elg and ReflectionUtils errors containing traversal sequences.
Conditions required:
  • Unauthenticated TCP/19009 access from Step 1
Where this breaks in practice:
  • None — the vulnerability is pre-auth with no prerequisites beyond network access
Detection/coverage: Inspect cpm.elg for usernames > 1,000 chars. Grep for ReflectionUtils errors containing ../ sequences. Check for concurrent FWM or MDS core dumps.
STEP 03

Malicious file upload to arbitrary path

Using the traversal primitive, the attacker uploads a malicious script or compiled Java class to an attacker-controlled filesystem path on the management server. The file upload mechanism in the management web service does not adequately validate the destination path, allowing placement of executable content outside the intended directory. This is a single-request operation combined with Step 2.
Conditions required:
  • Successful path traversal from Step 2
Where this breaks in practice:
  • None — file upload is part of the same exploit chain as the traversal
Detection/coverage: File integrity monitoring (FIM) on the management server filesystem. Watch for unexpected .sh, .py, .class, or .jar files in non-standard paths.
STEP 04

Arbitrary script execution and Java class loading

The management service executes the uploaded script or loads the attacker's Java class, granting code execution in the context of the CPM process. This is the culmination of the exploit chain — the attacker now has a shell on the most privileged management host in the Check Point infrastructure. Management services typically run with elevated privileges on Gaia OS.
Conditions required:
  • Malicious file placed in executable path from Step 3
Where this breaks in practice:
  • None — execution is triggered by the vulnerability mechanism itself
Detection/coverage: Process monitoring for unexpected child processes spawned by management services. Core dumps in /var/log/dump/usermode/. Unexpected outbound connections from the SMS.
STEP 05

Fleet compromise via management trust relationships

With code execution on the SMS/MDS, the attacker leverages built-in trust relationships to compromise the entire Check Point fleet. The SMS holds SIC certificates for every managed gateway, can push policy changes to all firewalls, and stores administrator credentials. The attacker can silently open firewall rules, disable logging, exfiltrate VPN configurations, create backdoor administrator accounts, and pivot to every managed gateway — all without triggering alerts on the gateways themselves, since policy pushes from the SMS are trusted by design.
Conditions required:
  • Code execution on the management server from Step 4
Where this breaks in practice:
  • None — the management server is designed to have full authority over all gateways
Detection/coverage: SmartConsole audit trail for unauthorized policy installations or administrator creation. Compare running gateway policies against last-known-good baselines. Monitor SIC certificate operations.
03 · Intelligence Metadata

The supporting signals.

In-the-Wild ExploitationConfirmed. Check Point observed a handful of pinpointed attacks on July 23, 2026 — two full months before public disclosure. CISA added to KEV on September 22, 2026 under BOD 26-04. No threat actor attribution disclosed.
Proof-of-Concept AvailabilityLimited. No fully public PoC on GitHub as of September 23, 2026. One reference noted by TheHackerWire. Threat actors clearly possess working exploits given confirmed in-the-wild use. Expect public PoCs within days given the pre-auth, low-complexity nature of the flaw.
EPSS ScoreNot yet computed. CVE was published September 22, 2026; FIRST.org EPSS model has not yet scored it. Given active exploitation and pre-auth RCE characteristics, expect a high percentile (>95th) once scored.
CISA KEV StatusListed. Added September 22, 2026. Remediation due date per BOD 26-04: September 25, 2026 (3 days). Forensic triage required. Note: the user's input indicated KEV: No, but CISA added this CVE on the same day as disclosure.
CVSS v3.1 VectorCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H — 9.8. Every base metric is worst-case except Scope (Unchanged). Network-attackable, low complexity, no auth, no user interaction, full CIA impact.
Affected VersionsSMS, MDS, Log Server, Multi-Domain Log Server, SmartEvent: R82.20, R82.10 <= Take 44, R82 <= Take 126, R81.20 <= Take 166, R81.10 <= Take 190 (EOS), all R80.x and R81 (EOS). LivePatch Take 28/29 does not fix this.
Fixed VersionsR82.20 Security Hotfix, R82.10 Take 45+, R82 Take 127+, R81.20 Take 170+, R81.10 Take 192+. No fix for EOS branches (R80.x, R81) — upgrade required. Details in SK1000171.
Exposure DataNo Shodan/Censys/GreyNoise counts published yet for this CVE specifically. Check Point community forums document recurring issues with TCP/19009 being exposed via implied rules and Trusted Clients: Any misconfigurations. Management servers *should* not be internet-facing but internal-network exposure is near-universal.
Disclosure TimelineExploitation observed: July 23, 2026. CVE published: September 20, 2026. Advisory updated and KEV listed: September 22, 2026. Approximately 61-day gap between first confirmed exploitation and public disclosure.
Related VulnerabilitiesSame advisory covers CVE-2026-85102 (actively exploited alongside this CVE) and related CVE-2026-16232 (SmartConsole auth bypass, CVSS 9.3, also exploited July 23, 2026). Same campaign, known attacker IPs: 151.241.99.207, 151.241.99.233, 158.62.198.182, 192.142.10.99, 139.28.37.250.
04 · The Call

Final Verdict
= UNCHANGED to CRITICAL (9.8/10)

Why this verdict

  • Pre-auth, zero-click, low complexity: The CVSS 9.8 vector (AV:N/AC:L/PR:N/UI:N) is fully accurate. No credentials, no user interaction, no chaining with another vulnerability required. A single HTTP request to TCP/19009 starts the chain. There is zero friction at the exploit level.
  • Active exploitation confirmed with KEV listing: Check Point confirmed targeted attacks on July 23, 2026 — two months before disclosure. CISA added this to the KEV catalog on September 22, 2026 with a 3-day remediation deadline under BOD 26-04. This is not theoretical risk; it is a confirmed zero-day that was exploited before any patch existed.
  • Role multiplier: The affected component IS the high-value role. The Check Point Security Management Server is definitionally a network-edge management plane — 100% of SMS/MDS installations serve as the central authority for all managed firewalls and VPN gateways. A successful chain yields: modification of all firewall policies fleet-wide, harvesting of VPN credentials and SIC certificates, silent disabling of logging and detection, and trusted-channel pivot to every managed gateway. Blast radius is fleet-scale. Per the high-value-role rule, when the affected component is canonically high-value (100% of installs occupy the role), and the outcome is fleet compromise, the verdict floor is CRITICAL.
  • Minimal real-world friction: While best practice says TCP/19009 should not face the internet, the management server is universally reachable from the internal network. Any post-initial-access attacker (phishing, VPN compromise, compromised workstation) can reach it. Check Point community forums document widespread Trusted Clients: Any misconfigurations and implied-rule exposure. Internal reachability does not meaningfully reduce severity for a management-plane component — it IS the target once you are inside.
  • No compensating detection at the exploit layer: There is no WAF, IPS, or EDR that typically inspects traffic to a Check Point management server's administrative port. The management web service handles its own authentication — which this CVE completely bypasses. Standard network security tools are blind to this attack path.

Why not higher?

CVSS 9.8 is already near the ceiling. The only metric preventing 10.0 is Scope: Unchanged, meaning the initial exploit is confined to the management server process context rather than crossing a security boundary to a different component. However, the post-exploitation blast radius (fleet-wide gateway compromise via trusted management channels) is not captured by CVSS base metrics. The operational severity arguably exceeds what 9.8 can express, but we cannot score above 10.0.

Why not lower?

Every downgrade argument fails under scrutiny. Internal-network-only exposure? The management server IS the crown jewel of the internal network, and 100% of deployments make it reachable from admin segments. Requires exploit chaining? No — this is a single-request, unauthenticated attack. Limited exploitation scope? Confirmed in the wild with CISA KEV listing and a 3-day federal deadline. Low install base? Check Point is one of the top three enterprise firewall vendors globally. There is no legitimate basis for reducing this below CRITICAL.

05 · Compensating Control

What to do — in priority order.

  1. Block TCP/19009 from all untrusted sources immediately — Create explicit firewall rules (NOT implied rules) restricting TCP/19009 access to named administrator IPs only. This is the single most effective mitigation and should be deployed within hours given active exploitation and KEV listing. Verify with netstat -tlnp | grep 19009 from the SMS and a port scan from an untrusted source. Per the noisgate mitigation SLA and the active-exploitation override: mitigate immediately, within hours.
  2. Apply the vendor hotfix for your branch now — Install R82.20 Security Hotfix, R82.10 Take 45+, R82 Take 127+, R81.20 Take 170+, or R81.10 Take 192+ per SK1000171. LivePatch Take 28/29 does NOT fix this — a full JHF or Security Hotfix install is required. The CISA KEV 3-day deadline (September 25, 2026) supersedes the noisgate CRITICAL remediation SLA of 90 days. Patch immediately.
  3. Hunt for indicators of compromise on every SMS/MDS — Check cpm.elg for usernames exceeding 1,000 characters. Grep for ReflectionUtils errors containing ../ traversal sequences. Look for concurrent FWM or MDS core dumps in /var/log/dump/usermode/. Review SmartConsole audit logs for unauthorized administrator creation or policy changes since July 23, 2026. Cross-reference source IPs 151.241.99.207, 151.241.99.233, 158.62.198.182, 192.142.10.99, 139.28.37.250 from the related campaign.
  4. Set Trusted Clients to explicit named IPs on all management servers — Change the Trusted Clients configuration from 'Any' to explicitly listed administrator IPs or ranges. The 'Any' setting is the primary reason management ports become internet-exposed or broadly reachable. This should be a permanent hardening measure, not just a temporary mitigation. Apply within hours alongside the port-blocking control.
  5. Upgrade all End-of-Support branches immediately — R80.x and R81 (non-point-release) have no fix available. These must be upgraded to a supported branch (R81.20+ or R82+) and then patched. There is no workaround for EOS versions beyond strict network isolation of TCP/19009. Begin the upgrade project immediately with a target completion within the noisgate CRITICAL remediation SLA of 90 days.
  6. Validate gateway policy integrity if any IoCs are found — If compromise indicators are detected, assume all managed gateways are affected. Compare running gateway policies against offline backups from before July 23, 2026. Rotate all SIC certificates, administrator credentials, and VPN pre-shared keys. Check for unauthorized administrator accounts, backdoor firewall rules, and unexpected policy installations.
What doesn't work
  • Check Point LivePatch (Take 28/29): Check Point explicitly states LivePatch does NOT address CVE-2026-93616. A full Jumbo Hotfix or Security Hotfix installation is required. Do not assume LivePatch provides protection.
  • IPS blade on the gateway fronting the SMS: Management traffic between SmartConsole and the SMS typically traverses implied rules that bypass IPS inspection. Even if IPS were inline, no signatures exist for this 1-day-old CVE.
  • WAF or reverse proxy in front of the management server: Management servers do not sit behind WAFs. The management web service on TCP/19009 uses a proprietary protocol stack, not standard HTTP that a WAF can parse or filter.
  • Network-level rate limiting or throttling: This is a single-request exploit. A one-shot attack completes before any rate limiter would trigger.
06 · Verification

Crowdsourced verification payload.

Run this script directly on each Check Point Management Server (SMS or MDS) in Expert mode as root. SSH to the appliance, enter expert mode, then execute: bash check_cve_2026_93616.sh. No arguments required. The script checks the installed Gaia version and JHF take level against the fixed versions and outputs PATCHED, VULNERABLE, or UNKNOWN with appropriate exit codes.

noisgate-verify.sh
BASHREAD-ONLYSAFE
#!/bin/bash
# CVE-2026-93616 - Check Point Management Server Path Traversal
# Run on the SMS/MDS appliance in Expert mode as root.
# Exits: 0=PATCHED  1=VULNERABLE  2=UNKNOWN

echo "=== CVE-2026-93616 Vulnerability Check ==="
echo "Check Point Management Server Pre-Auth Path Traversal"
echo ""

[ ! -f /etc/cp-release ] && echo "UNKNOWN - not a Check Point appliance" && exit 2
REL=$(cat /etc/cp-release)
echo "Release: $REL"

cpprod_util FwIsFirewallMgmt 2>/dev/null | grep -q 1 || { echo "UNKNOWN - not a management server"; exit 2; }
echo "Role: Management Server confirmed"

TAKE=0
for f in /opt/CPsuite-*/fw1/conf/jumbo_hf_installed_take.conf; do
  [ -f "$f" ] && TAKE=$(tr -dc '0-9' < "$f") && break
done
echo "Jumbo Hotfix Take: $TAKE"
echo ""

check() {
  if [ "$TAKE" -ge "$1" ] 2>/dev/null; then
    echo "PATCHED - Take $TAKE (fixed in Take $1+)"
    exit 0
  else
    echo "VULNERABLE - Take $TAKE (needs Take $1 or later)"
    exit 1
  fi
}

case "$REL" in
  *R82.20*)
    cpinfo -y all 2>/dev/null | grep -qi "Security Hotfix" && echo "PATCHED - R82.20 Security Hotfix applied" && exit 0
    echo "VULNERABLE - R82.20 missing Security Hotfix"
    exit 1;;
  *R82.10*) check 45;;
  *R82*)    check 127;;
  *R81.20*) check 170;;
  *R81.10*) check 192;;
  *R80*|*R81*)
    echo "VULNERABLE - $REL is End of Support with no fix. Upgrade required."
    exit 1;;
  *)
    echo "UNKNOWN - unrecognized version"
    exit 2;;
esac
07 · Sources

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.