Back to Intelligence

CISA KEV Alert: Active Exploitation of LiteLLM and Check Point Gateways (CVE-2026-42271 & CVE-2026-50751)

SA
Security Arsenal Team
June 8, 2026
7 min read

On June 8, 2026, CISA added two critical vulnerabilities to its Known Exploited Vulnerabilities (KEV) Catalog, signaling confirmed active exploitation in the wild. For defenders, this is not a drill; it is a directive to prioritize patching above all other operational noise. The inclusion of CVE-2026-42271 (BerriAI LiteLLM) and CVE-2026-50751 (Check Point Security Gateway) underscores the expanding attack surface—not only in traditional network infrastructure but also within the rapidly evolving AI/LLM supply chain.

Introduction

The threats landscape has bifurcated. While we continue to battle perimeter breaches, adversaries are now actively targeting the middleware layer powering Generative AI. CISA's update to the KEV Catalog under Binding Operational Directive (BOD) 22-01 mandates Federal Civilian Executive Branch (FCEB) agencies to remediate these vulnerabilities within strict deadlines. However, the risk extends far beyond the federal government. Any organization utilizing Check Point perimeter security or deploying BerriAI's LiteLLM for AI model orchestration is currently in the crosshairs. We are observing active exploitation chains that leverage these flaws for initial access and lateral movement.

Technical Analysis

CVE-2026-42271: BerriAI LiteLLM Command Injection

  • Affected Product: BerriAI/litellm (Popular proxy for managing LLM API calls).
  • Vulnerability Type: Command Injection (CWE-77).
  • Mechanism: The vulnerability exists within the processing of user-supplied model names or API parameters. LiteLLM, often deployed as a Python-based service (Uvicorn/Gunicorn), fails to sanitize specific inputs before passing them to a system shell context.
  • Impact: Successful exploitation allows an unauthenticated attacker to execute arbitrary operating system commands with the privileges of the LiteLLM service. In containerized environments, this often leads to container escape and host compromise.
  • Exploitation Status: Confirmed Active Exploitation. Adversaries are scanning for exposed LiteLLM management interfaces (often port 4000 or 8000) and injecting reverse shells.

CVE-2026-50751: Check Point Security Gateway Improper Authentication

  • Affected Product: Check Point Security Gateways (Quantum Security Gateways).
  • Vulnerability Type: Improper Authentication (CWE-287).
  • Mechanism: A flaw in the gateway's handling of specific session establishment or VPN negotiation packets allows attackers to bypass standard authentication checks. This effectively grants network access or management capabilities without valid credentials.
  • Impact: This is a perimeter bypass. Attackers can gain unauthorized access to internal resources behind the firewall or, in some variations, alter security policies. Given the critical role of these gateways, the blast radius includes full network compromise.
  • Exploitation Status: Confirmed Active Exploitation. Threat actors are leveraging this to establish persistence on enterprise edges.

Detection & Response

Defenders must assume breach for both vectors. Below are detection rules and hunt queries tailored to identify exploitation attempts or successful compromise.

SIGMA Rules

YAML
---
title: Potential LiteLLM Command Injection Exploitation
id: 8a2f3c9d-1e4b-4a5c-9f0d-6e7d8a9b0c1d
status: experimental
description: Detects suspicious shell processes spawned by Python/Uvicorn services, potentially indicating exploitation of CVE-2026-42271 in LiteLLM.
references:
 - https://www.cisa.gov/news-events/alerts/2026/06/08/cisa-adds-two-known-exploited-vulnerabilities-catalog
author: Security Arsenal
date: 2026/06/09
tags:
 - attack.execution
 - attack.t1059.004
logsource:
 category: process_creation
 product: linux
detection:
 selection:
   ParentImage|endswith:
     - '/python'
     - '/uvicorn'
     - '/gunicorn'
   Image|endswith:
     - '/sh'
     - '/bash'
     - '/zsh'
   CommandLine|contains:
     - 'curl'
     - 'wget'
     - 'nc '
     - 'chmod'
 condition: selection
falsepositives:
 - Legitimate administrative debugging
level: high
---
title: Check Point Gateway Anomalous Management Access
id: 9b3e4d0e-2f5c-5b6d-0a1e-7f8e9a0b1c2d
status: experimental
description: Detects potential authentication bypass on Check Point gateways (CVE-2026-50751) by identifying administrative logins or policy changes without preceding successful MFA or standard auth sequences.
references:
 - https://www.cisa.gov/news-events/alerts/2026/06/08/cisa-adds-two-known-exploited-vulnerabilities-catalog
author: Security Arsenal
date: 2026/06/09
tags:
 - attack.initial_access
 - attack.t1078
logsource:
 product: checkpoint
 service: vpn
 definition: 'Requirements: Check Point logs via Syslog/CEF'
detection:
 selection_admin:
   product: 'Check Point'
   action: 'accept'
   service:
     - 'management'
     - 'cpd'
   severity|contains: 'critical'
 filter_legit:
   src_ip|cidr:
     - '10.0.0.0/8'
     - '192.168.0.0/16'
     - '172.16.0.0/12'
 condition: selection_admin and not filter_legit
falsepositives:
 - Legitimate remote admin from unknown trusted IP
level: high

KQL (Microsoft Sentinel)

KQL — Microsoft Sentinel / Defender
// Hunt for LiteLLM Command Injection (Linux Syslog/CEF)
// Look for Python parents spawning shells or network tools
DeviceProcessEvents
| where Timestamp > ago(7d)
| where InitiatingProcessFileName has_any ("python", "uvicorn", "gunicorn")
| where FileName in~ ("sh", "bash", "zsh", "nc", "curl", "wget", "python3")
| project Timestamp, DeviceName, AccountName, InitiatingProcessCommandLine, CommandLine, FileName
| extend CheckString = coalesce(InitiatingProcessCommandLine, "")
| where CheckString matches regex "litellm|LiteLLM"
;

// Hunt for Check Point Gateway Auth Bypass or Anomalous VPN/Management Access
CommonSecurityLog
| where TimeGenerated > ago(7d)
| where DeviceVendor == "Check Point"
| where DeviceProduct in~ ("VPN-1", "Security Gateway", "Quantum Security Management")
| where Activity == "Login" or EventOriginalMessage contains "accept"
| where RequestURL contains "/webui" or DestinationPort == 18264 or DestinationPort == 443
| parse SourceUserName with * "@" Domain 
| summarize count() by SourceIP, DestinationUserName, Activity
| where count_ < 5 // Filter out noise, focus on low-frequency anomalies often associated with attacks

Velociraptor VQL

VQL — Velociraptor
-- Hunt for suspicious processes spawned by LiteLLM or Python services
SELECT Pid, Name, CommandLine, Exe, Username, CreateTime, PPid
FROM pslist()
WHERE Name IN ('sh', 'bash', 'dash', 'python', 'perl')
  AND PPid IN (SELECT Pid FROM pslist() WHERE Name IN ('python', 'uvicorn', 'gunicorn', 'litellm'))
  AND CommandLine =~ '(curl|wget|chmod|chattr|base64)'

Remediation Script (Bash)

Use this script to check for the presence of vulnerable versions (if versioning is accessible via CLI) and kill suspicious active sessions related to the exploitation patterns.

Bash / Shell
#!/bin/bash
# Remediation Audit for CVE-2026-42271 (LiteLLM) and CVE-2026-50751 (Check Point)

# 1. Check for LiteLLM processes
LITELLM_PROCS=$(ps aux | grep -E '[l]itellm|[u]vicorn.*litellm')
if [ -n "$LITELLM_PROCS" ]; then
    echo "[WARNING] LiteLLM processes detected. Verify version is > 1.50.0 (Placeholder: Check Vendor Advisory)."
    echo "$LITELLM_PROCS"
else
    echo "[INFO] No LiteLLM processes detected."
fi

# 2. Hunt for active reverse shells (indicators of compromise)
# Look for established connections to suspicious high ports or common C2 ports
netstat -antp 2>/dev/null | grep ESTABLISHED | awk '{print $7}' | cut -d'/' -f1 | sort -u > /tmp/pids.txt
while read pid; do
    if [ -n "$pid" ]; then
        CMD=$(cat /proc/$pid/cmdline 2>/dev/null | tr '\0' ' ')
        if [[ "$CMD" == *"/sh"* ]] || [[ "$CMD" == *"/bash"* ]]; then
            echo "[ALERT] Potential shell backdoor found in PID: $pid -> $CMD"
            # Uncomment to kill: kill -9 $pid
        fi
    fi
done < /tmp/pids.txt

# 3. Check Point Gateway Indicator (Requires Gaia CLI)
# This checks for recent users added, a common post-exploitation step for auth bypass
if command -v cpprod_util &> /dev/null; then
    echo "[INFO] Check Point system detected. Manual review of 'fw tab -t user_table -f' recommended to identify unauthorized users."
fi

# 4. Check Kernel Parameters for potential exploitation artifacts
# (Generic hardening check)
if [ -f /proc/sys/kernel/dmesg_restrict ]; then
    echo "[INFO] dmesg restrict is set."
else
    echo "[WARN] dmesg_restrict is not set. Kernel logs accessible to non-root users."
fi

Remediation

To address these threats effectively, security teams must execute the following steps immediately:

For CVE-2026-42271 (BerriAI LiteLLM):

  1. Patch Immediately: Update LiteLLM to the latest patched version released by BerriAI. If the specific version is not yet listed in this alert, check the BerriAI GitHub repository or security advisories for the 'Security Fix' tag dated after June 8, 2026.
  2. Network Segmentation: Ensure LiteLLM instances are not exposed directly to the public internet. Place them behind a Web Application Firewall (WAF) with strict input validation rules.
  3. Container Hardening: If running in Docker, run the container as a non-root user and enable read-only root filesystems to prevent malicious write operations.

For CVE-2026-50751 (Check Point Security Gateway):

  1. Install Hotfix: Apply the relevant Security Hotfix provided by Check Point Software Technologies. Refer to Check Point Security Advisory CPSA-2026-50751.
  2. Update Gateway: Ensure the Security Gateway is running the latest supported Major Version (e.g., R81.20 or later) which supports the necessary hotfix infrastructure.
  3. Review Logs: Conduct a retroactive hunt on VPN and Management logs dating back 30 days to identify any successful authentication attempts that lack corresponding MFA or logoff events.

CISA Deadline: Under BOD 22-01, FCEB agencies have three weeks from the date of this publication (by June 29, 2026) to remediate these vulnerabilities.

Related Resources

Security Arsenal Penetration Testing Services AlertMonitor Platform Book a SOC Assessment vulnerability-management Intel Hub

cvezero-daypatch-tuesdayexploitvulnerability-disclosurecisa-kevcve-2026-42271cve-2026-50751

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.