SecPod

Learn Search

Search across all Learn content

← Back to Security Research
Check Point Zero-Day Attacks Deploy TSGA Backdoor, TSProxy Malware, and Destructive Wiper

Check Point Zero-Day Attacks Deploy TSGA Backdoor, TSProxy Malware, and Destructive Wiper

Nov 16, 2026By Manasvini R
malware researchthreat intelligence

Summary

Check Point Research identified an active exploitation campaign beginning October 30, 2026, targeting vulnerable Check Point Security Gateways. The initial attacks focused on organizations in Israel, followed by a broader wave on November 2 targeting organizations across multiple countries and industries.

Attackers exploited PHP2Shell vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, to gain initial access through exposed Mobile Access portals. They subsequently exploited CVE-2026-123039 in the Gaia OS DynamicCLI interface to elevate privileges and execute commands as root.

Following compromise, the attackers installed PHP webshells, deployed the TSGA persistent backdoor and embedded TSProxy networking component, and used the gogo scanner to identify internal assets. In one investigated intrusion, attackers moved laterally to a VMware ESXi host over SSH and deployed the destroy_v6 wiper, causing destructive file deletion.

Check Point reported active exploitation affecting approximately two dozen customers. Researchers also identified similarities between TSGA, TSProxy, and the PhiliKit malware framework previously associated with attacks against Ivanti appliances. However, the available evidence does not establish definitive attribution to a particular threat actor.

Background

Check Point Security Gateways are network security appliances that provide firewall, VPN, intrusion prevention, and remote access capabilities. Their Mobile Access portals allow authorized users to connect remotely to enterprise applications and internal resources. Because these gateways operate at the network perimeter, successful exploitation can provide attackers with a foothold into an organization's internal infrastructure.

The DynamicCLI daemon, known as dynamic_server, runs with root privileges and listens on a local Unix socket at /tmp/dynamic_socket. Insufficient authentication of requests sent through this world-writable socket allows lower-privileged processes to execute commands with elevated privileges.

Following successful exploitation, attackers deployed multiple malware components, including PHP webshells, the TSGA backdoor, and its embedded TSProxy networking component. These tools enabled persistent remote access, command execution, covert communication, and further compromise of internal systems.

TSGA is a Linux-based persistent backdoor specifically adapted to Check Point Security Gateways. It provides encrypted remote access, command execution, configuration theft, file transfers, process injection, and persistence capabilities. The malware can also interfere with firewall logs, making it more difficult for defenders to identify malicious activity.

TSProxy is a shared-library backdoor embedded within TSGA. It supports encrypted SOCKS-style proxy communications and can be injected into existing appliance processes. This enables attackers to route malicious traffic through compromised systems while blending with legitimate network activity.

Check Point Research identified significant technical similarities between TSGA, TSProxy, and PhiliKit, a previously documented malware framework targeting Ivanti VPN appliances. PhiliKit has been associated with the China-nexus threat group UNC5221. However, researchers have not conclusively attributed the Check Point attacks to this group.

During the observed attacks, exploitation activity originated from attacker-associated infrastructure, including 130.94.19[.]204. Researchers identified suspicious PHP session artifacts, repeated Apache/PHP worker crashes, and malicious files on compromised gateways. These findings demonstrate how attackers combined vulnerability exploitation with persistent backdoors and destructive malware to expand their access beyond the initially compromised security appliances.

Vulnerability Details

CVE ID CVSS Score EPSS Score Affected Product Vulnerability Type
CVE-2026-123039 7.8 (High) 2.19% Check Point Gaia OS on affected Security Gateway and Management platforms Local Privilege Escalation
CVE-2026-109618 9.2 (Combined PHP2Shell assessment) 100% Check Point Security Gateway Mobile Access and affected PHP-based interfaces Remote Code Execution
CVE-2026-109619 9.2 (Combined PHP2Shell assessment) 100% Check Point Security Gateway Mobile Access and affected PHP-based interfaces Remote Code Execution

Attack Methodology

Phase 1: Initial Access Through PHP2Shell

The attackers began by targeting internet-facing Check Point Mobile Access portals using the PHP2Shell vulnerabilities, CVE-2026-109618 and CVE-2026-109619. These flaws enabled unauthorized code execution on vulnerable appliances.

During exploitation, PHP session files were created under /opt/CPcvpn-R81.20/tmp/, including artifacts containing the marker upload_progress_pwn. Investigators also observed hundreds of segmentation faults affecting Apache/PHP workers, indicating repeated exploitation attempts against the exposed portal.

Phase 2: Root Privilege Escalation Through CVE-2026-123039

After obtaining an initial foothold, attackers exploited CVE-2026-123039 in the Gaia OS DynamicCLI interface to escalate privileges.

The dynamic_server daemon executes as root and accepts command requests through the world-writable Unix socket /tmp/dynamic_socket. Because the interface does not adequately authenticate connecting processes, an attacker with local access can submit commands for execution with elevated privileges.

This enabled the attackers to move from the lower-privileged Mobile Access portal service account to root-level command execution, providing the permissions necessary to modify protected system files and install persistent malware.

Phase 3: PHP Webshell Deployment

Using their elevated privileges, the attackers created malicious PHP template files within the Mobile Access portal's template directory. The webshell filenames followed the pattern zz*.tpl.

The attackers initially installed lightweight command-execution webshells and subsequently replaced them with a more capable version protected by the hardcoded authentication key zz2026.

The advanced webshell supported operating-system command execution, PHP code evaluation, and file uploads. It could process plain-text requests as well as Base64-encoded and XOR-obfuscated commands. Uploaded files were written to /tmp/.up.tpl.

The attackers also used the filename challenge_render.tpl to conceal malicious functionality within the portal's normal template structure.

Phase 4: TSGA Backdoor Installation and Persistence

After deploying the webshells, the attackers installed the TSGA backdoor, initially launched from /tmp/CPconfigdaemon. The malware redirected its output to the hidden log file /tmp/.bd.log.

Once executed with root privileges, TSGA relocated itself into a boot persistence directory and relaunched under the filename slot_CPconfigdaemon, enabling continued access after system restarts.

The malware provided encrypted remote command execution, interactive shell access, file transfers, appliance configuration collection, backup and snapshot-based persistence, process injection, and binary firewall-log tampering.

TSGA also supported authenticated communications over TCP port 11072 and the Unix socket /tmp/.p_sock. Researchers determined that its networking and command-execution functionality shared significant code similarities with the previously documented PhiliKit malware framework.

Phase 5: TSProxy Deployment and Covert Communication

The TSGA implant contained an embedded shared-library backdoor known as TSProxy. The malware extracted the library to /lib/libcpfwcommon.so and injected it into selected appliance processes.

TSProxy provided encrypted SOCKS-style proxy functionality, allowing attackers to route communications through the compromised gateway.

By intercepting networking functions inside legitimate processes, the implant could recognize specially marked attacker connections while allowing ordinary network traffic to continue.

Check Point researchers identified matching credentials, transport secrets, activation-selector digests, and cryptographic certificate material between TSProxy and the previously identified PhiliKit component libdssafe.so, demonstrating a strong relationship between the malware families.

Phase 6: Internal Network Reconnaissance

With persistent access established, the attackers deployed the gogo network scanner to discover additional systems and services within the compromised organization's internal network.

The scanner supported identification of reachable assets and exposed services, providing information that could be used to identify systems for further compromise.

Using the compromised gateway as an entry point, the attackers subsequently initiated lateral movement toward internal infrastructure.

Phase 7: VMware ESXi Lateral Movement

In one investigated intrusion, the attackers accessed a VMware ESXi host from the compromised Check Point gateway using SSH.

The activity involved the vpxuser service account, which is normally used for communication between VMware vCenter and managed ESXi hosts.

After gaining access to the virtualization infrastructure, the attackers staged additional payloads and prepared a destructive operation targeting files on the ESXi system.

Phase 8: Destructive Wiper Deployment

The attackers deployed a Linux ELF wiper named destroy_v6 onto the compromised ESXi environment.

The wiper recursively traversed specified directories and deleted files regardless of their extensions. By default, it attempted to overwrite file contents with zeros before removing them.

The malware also supported a fast-deletion option that skipped overwriting and an exclusion mechanism that allowed operators to specify directories that should not be processed.

During execution, the wiper created a politically themed note named DREAM_ANNEXATION.txt in processed directories. The note contained a timestamp and rejected ransom negotiations, indicating a destructive objective rather than conventional ransomware extortion.

Visual Attack Flow

01   ENTRY POINT
Internet-facing Check Point Security Gateway

Mobile Access portal exposed to remote requests

02   INITIAL EXPLOITATION
PHP2Shell remote code execution

CVE-2026-109618 and CVE-2026-109619

03   PRIVILEGE ESCALATION
Gaia DynamicCLI root access

CVE-2026-123039

04   PERSISTENCE
Malicious PHP webshells

zz*.tpl and challenge_render.tpl

05   BACKDOOR DEPLOYMENT
TSGA and TSProxy

slot_CPconfigdaemon and /lib/libcpfwcommon.so

06   RECONNAISSANCE
Internal discovery using gogo

Identify reachable internal assets and services

07   LATERAL MOVEMENT
SSH access to VMware ESXi

Observed in an investigated intrusion

08   DESTRUCTIVE ACTION
destroy_v6 wiper deployed

File deletion and DREAM_ANNEXATION.txt note

Indicators of Compromise (IoCs)

Check Point Research published network indicators, malicious file hashes, and host-based artifacts associated with the investigated attacks. The following indicators can assist security teams in identifying potential exploitation, malware deployment, and post-compromise activity.

Network Indicators

Indicator Description
130.94.19[.]204 Attacker-associated infrastructure observed targeting Check Point Mobile Access portals during the initial exploitation wave
130.94.19[.]168 Additional attacker-associated infrastructure listed in Check Point Research's IoC table

File Indicators

Filename / Malware SHA-256
PHP .tpl Webshell edb3e92eb7bda7413a09edcc4b1ba3cbed4bcdf42808a8abce63b6e4c3fda2b3
slot_CPconfigdaemon (TSGA) 4171b521fa6441382aa2fa014a08e18ff9a6e0ab6096c9144072bf46b37a2ec8
CPconfigdaemon (TSGA) d7549671c7189979f0624ec52dad9ea6a707d23146e3585ab8f8c4720eca0c63
gogo Scanner 3b51080a909023f7d484e2443740b572307ce0054eb9e32d98f0a06bfcd2db90
destroy_v6 Wiper c7d9477309fac6c638a1128442056a242de086996add8cb49cd6496c36003237
DREAM_ANNEXATION.txt 0522937c0c0337be54fe25493539e5f5f8ceb9742c980aaff1cfc9515f3a96f7

Related PhiliKit Malware Indicators

The following hashes were also published in Check Point Research's IoC table as samples from the related PhiliKit malware framework. They provide useful historical and malware-family context but should not automatically be treated as files deployed during the investigated Check Point intrusion.

Filename SHA-256
philistine ee16e1695711b9d60c62925fcfb2fe1a29c18ae26565bc05410a14e16d2a39d7
libdssafe.so fb1a6235eb6bce8c9ec8cd8bb85f17c9c2bde8ecb45b7a243a30ceadf26654da

Host-Based Indicators

Indicator Description
/opt/CPcvpn-R81.20/tmp/ Observed PHP session artifact location associated with initial exploitation
upload_progress_pwn Marker found in suspicious PHP session files
zz*.tpl Malicious PHP template filename pattern
challenge_render.tpl Additional webshell filename used in the Mobile Access portal
zz2026 Hardcoded key associated with the advanced PHP webshell
/tmp/.up.tpl Webshell file-upload destination
/tmp/CPconfigdaemon Initial TSGA backdoor deployment path
slot_CPconfigdaemon TSGA persistence executable filename
/tmp/.bd.log Hidden output log associated with TSGA execution
/lib/libcpfwcommon.so TSProxy shared-library backdoor
/tmp/.p_sock Unix socket used by the backdoor communication framework
TCP 11072 Backdoor communication port observed in the TSGA/PhiliKit tooling
destroy_v6 Linux ELF destructive wiper
DREAM_ANNEXATION.txt Note created during the destructive operation

Investigators should also examine /var/log/dynamicCli.log for unusual privileged command execution, unexpected use of the system() execution sink, and commands bypassing the expected Clish wrapper. Repeated Apache/PHP segmentation faults and unexplained firewall-log gaps may provide additional evidence when correlated with the indicators above.

Some paths, filenames, ports, and commands are not inherently malicious. Correlate these indicators with process activity, timestamps, network telemetry, and other forensic evidence before confirming compromise.

Mitigation

Organizations operating Check Point Security Gateways and affected Gaia OS systems should implement the following security measures to prevent exploitation and investigate possible compromise.

  1. • Apply Check Point Security Hotfixes: Install the latest vendor-recommended hotfixes and Jumbo Hotfix packages for the affected Gaia OS and Security Gateway releases. Address both the PHP2Shell vulnerabilities and CVE-2026-123039.
  2. • Secure Internet-Facing Mobile Access Portals: Identify exposed Mobile Access services, review their configurations, and restrict unnecessary access according to Check Point's official guidance.
  3. • Inspect Gateway Files and Processes: Search for unexpected PHP templates, suspicious session files, slot_CPconfigdaemon, CPconfigdaemon, and libcpfwcommon.so.
  4. • Protect Virtualization Infrastructure: Restrict privileged SSH access, monitor ESXi systems for unauthorized file modifications, and maintain verified, isolated backups to support recovery from destructive attacks.

Instantly Fix Risks in Endpoints & Cloud with Saner

Saner is a continuous and integrated vulnerability & exposure management platform that instantly fixes risks in endpoints and cloud.

Featured Posts

Open CVE-2026-108186 and CVE-2026-108187: Zoom Workplace Flaws Enable Remote Privilege Escalation
CVE-2026-108186 and CVE-2026-108187: Zoom Workplace Flaws Enable Remote Privilege Escalation

CVE Research

CVE-2026-108186 and CVE-2026-108187: Zoom Workplace Flaws Enable Remote Privilege Escalation

Nov 16, 2026

Open November 2026 Patch Wave: Major Security Fixes Across Apple, Adobe, Chrome, and Firefox
November 2026 Patch Wave: Major Security Fixes Across Apple, Adobe, Chrome, and Firefox

CVE Research

November 2026 Patch Wave: Major Security Fixes Across Apple, Adobe, Chrome, and Firefox

Highlight Chrome's confirmed in-the-wild exploitation of CVE-2026-125350, followed by the critical security fixes across Apple, Adobe, and Firefox.

Nov 13, 2026

Open ScreenConnect Attacks Deploy SCKit Malware Across MSP Environments, Targeting 288 Endpoints in 48 Organizations
ScreenConnect Attacks Deploy SCKit Malware Across MSP Environments, Targeting 288 Endpoints in 48 Organizations

CVE Research

ScreenConnect Attacks Deploy SCKit Malware Across MSP Environments, Targeting 288 Endpoints in 48 Organizations

A new worm-like malware toolkit called SCKit targets managed service providers (MSPs) by abusing ScreenConnect remote access sessions. The campaign attempts to move from infected customer endpoints to MSP technicians' workstations, potentially exposing additional customer networks. Researchers identified activity affecting at least 288 endpoints across 48 organizations.

Nov 13, 2026

Open Critical Apache HTTP/2 RCE Flaw Lets Unauthenticated Attackers Take Full Control of the Server
vulnerability researchCritical Apache HTTP/2 RCE Flaw Lets Unauthenticated Attackers Take Full Control of the Server

CVE Research

Critical Apache HTTP/2 RCE Flaw Lets Unauthenticated Attackers Take Full Control of the Server

CVE-2026-119618 is a critical race-condition vulnerability in Apache HTTP Server’s mod_http2 that can lead to unauthenticated remote code execution when HTTP/2 is enabled. Versions 2.4.55 through 2.4.70 are affected and fixed in 2.4.71. Only hosts that negotiate h2 or h2c are exposed. This article covers how the flaw works, distribution fixes, temporary Protocols mitigation, and related issues in the same release.

Nov 12, 2026