# Chinese APT Groups (HOLLOW/VAULT/GENESIS/MURKY/VERTIGO PANDA) Conduct Cyberespionage Against Global Banking Networks

> CrowdStrike's 2026 Financial Services Threat Landscape Report and Team Cymru infrastructure data describe China-nexus adversaries (HOLLOW, VAULT, GENESIS, MURKY and VERTIGO PANDA) targeting banks, fintechs and financial services in South America, Southeast Asia, the Middle East and North America for long-term intelligence collection. Reported tooling includes ShadowPad, Vshell, KEYPLUG and Fscan, plus an ORB network used to reach Microsoft 365 mailboxes.

- **Published:** 2026-11-18T09:45:10Z
- **Last reviewed:** 2026-11-18T09:45:10Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-6321
- **Card image:** https://intel.threadlinqs.com/og/threat/TL-2026-6321.png
- **ID:** TL-2026-6321
- **Severity:** HIGH
- **Category:** APT
- **Status:** ACTIVE
- **Actor:** HOLLOW PANDA (China)
- **Detections:** 9 (1 shown free below) · **IOCs:** 14

## Description

GBHackers (2026-11-18), citing CrowdStrike and Team Cymru, reports a shift by China-nexus actors from financially motivated theft toward long-term cyberespionage against financial institutions. The CrowdStrike 2026 Financial Services Threat Landscape Report (published 2026-05-14; reporting period 2025-04-01 to 2026-03-31) states that China-nexus adversaries pose the most significant intelligence-collection threat to financial services organizations, especially in South and Southeast Asia, and describes shared TTPs: exploitation of edge devices, DLL search-order hijacking, use of compromised infrastructure for command and control, and targeting of cloud environments.

Per the reporting: HOLLOW PANDA targets financial institutions in South America and Southeast Asia, exploits Check Point VPN appliances and deploys the ShadowPad backdoor (active since at least 2023; objectives are intelligence gathering and IP theft). VAULT PANDA targets financial institutions and supporting entities across multiple regions and deploys KEYPLUG via DLL search-order hijacking. GENESIS PANDA targeted a Southeast Asia-based financial entity and a North American fintech using VShell implants and FScan utilities, with infrastructure linked to earlier China-nexus operations. MURKY PANDA operates a unique Chinese operational relay box (ORB) network to access Microsoft 365 email accounts from more than 150 IP addresses in 36 countries; financial services were among its most frequently targeted sectors out of 340 organizations across 30+ sectors. VERTIGO PANDA is named in the GBHackers article but no details are provided in the available sources. The stated collection goals are economic data and PII.

Team Cymru observed 64 servers matching known ShadowPad C2 patterns and 115 servers matching Vshell patterns between 2026-01-01 and 2026-06-01, and found one Vshell-tagged server with an open directory containing files named 'fscan' and 'vs' (Fscan and Vshell binaries). Team Cymru did not attribute these server sets to any specific actor, and the counts measure infrastructure, not compromised institutions. No CVE is named for the Check Point VPN exploitation, and no IP addresses, domains or hashes are published in the article text; network IOCs are therefore not included. Severity is analyst-assigned (no CVSS).

## MITRE ATT&CK

- T1190 Exploit Public-Facing Application
- T1574.001 Hijack Execution Flow: DLL
- T1574.001 Hijack Execution Flow: DLL
- T1574.001 Hijack Execution Flow: DLL
- T1090.003 Proxy: Multi-hop Proxy
- T1046 Network Service Discovery
- T1114.002 Email Collection: Remote Email Collection
- T1584 Compromise Infrastructure

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**TL-2026-6321 Check Point VPN Gateway Originating Outbound Sessions to External Hosts (HOLLOW PANDA / ShadowPad Staging)** (`TL-2026-6321-SPL-001`)

HOLLOW PANDA exploits Check Point VPN appliances for initial access (no CVE named in CrowdStrike reporting) and then deploys ShadowPad. A compromised gateway commonly begins originating its own sessions to adversary staging or C2 hosts. This search is behaviour-first: it flags accepted Check Point sessions where the source is the enforcement gateway itself (src equals orig) and the destination is a non-RFC1918 address, excluding DNS and NTP, and alerts when the gateway talks to three or more distinct external hosts in 24 hours. Mapped to T1133 because the VPN gateway is the abused external remote service, and to T1105 because gateway-originated sessions to new external hosts are the typical ingress-tool-transfer stage that stages ShadowPad.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1190, T1133, T1105
- **Data sources:** checkpoint_firewall_logs, index=checkpoint sourcetype=cp_log, CommonSecurityLog
- **Severity:** high
- **Confidence:** medium

### Splunk SPL

````spl
index=checkpoint sourcetype=cp_log action=Accept earliest=-24h
```HOLLOW PANDA Check Point VPN exploitation: a gateway should rarely originate sessions to several distinct external hosts```
| where src=orig
| where NOT (cidrmatch("10.0.0.0/8",dst) OR cidrmatch("172.16.0.0/12",dst) OR cidrmatch("192.168.0.0/16",dst))
| where NOT match(lower(coalesce(service_id,service,"")),"^(domain-udp|domain-tcp|ntp-udp|ntp|53|123)$")
| stats count as connections dc(dst) as unique_destinations values(dst) as destinations values(service_id) as services min(_time) as first_seen max(_time) as last_seen by orig
| where unique_destinations>=3
| convert ctime(first_seen) ctime(last_seen)
````

### Microsoft KQL

```kql
CommonSecurityLog
| where TimeGenerated > ago(24h)
| where DeviceVendor =~ "Check Point"
| where DeviceAction in~ ("Accept", "Allow")
| where SourceIP == DeviceAddress
| where not(ipv4_is_private(DestinationIP))
| where DestinationPort !in (53, 123)
| summarize Connections = count(), UniqueDestinations = dcount(DestinationIP), Destinations = make_set(DestinationIP, 25), Ports = make_set(DestinationPort, 15), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated) by DeviceAddress, DeviceProduct
| where UniqueDestinations >= 3
```

### Sigma

```yaml
title: Check Point VPN Gateway Originating Outbound Sessions To External Hosts
id: 3c0f6a52-8d1b-4e7a-b6f4-2a9d51c7e801
status: experimental
description: Detects accepted Check Point sessions where the source is the enforcement gateway itself (src equals orig) and the destination is a non-RFC1918 address. HOLLOW PANDA exploits Check Point VPN appliances for initial access and a compromised gateway may call out to staging or C2 infrastructure (ShadowPad delivery). Aggregate by orig and alert on three or more distinct destinations.
references:
  - https://www.crowdstrike.com/en-us/blog/crowdstrike-2026-financial-services-threat-landscape-report
  - https://gbhackers.com/global-banking-networks/
author: Threadlinqs Team
date: 2026-11-18
tags:
  - attack.initial-access
  - attack.t1190
  - attack.persistence
  - attack.t1133
  - attack.command-and-control
  - attack.t1105
  - attack.s0596
logsource:
  product: checkpoint
  service: firewall
detection:
  selection:
    action: 'Accept'
    src|fieldref: orig
  filter_private_dst:
    dst|cidr:
      - '10.0.0.0/8'
      - '172.16.0.0/12'
      - '192.168.0.0/16'
  filter_infra_services:
    service_id:
      - 'domain-udp'
      - 'domain-tcp'
      - 'ntp-udp'
  condition: selection and not 1 of filter_*
falsepositives:
  - Gateway software updates, licensing, CRL/OCSP and ThreatCloud lookups to vendor infrastructure
  - Administrator troubleshooting (ping, traceroute, curl) run from the gateway
level: high

```

**Known false positives:**

- Gateway software updates, licensing, CRL/OCSP and ThreatCloud lookups to Check Point or vendor infrastructure
- Administrator troubleshooting (ping, traceroute, curl) executed from the gateway shell
- Site-to-site VPN or SD-WAN tunnel establishment where the gateway is the session originator

**Tuning notes:** T1133/T1105 coverage: review the destinations field against threat-intel and ASN data; first-seen external hosts contacted by the gateway are the strongest sign of payload staging. Field names assume the Check Point Splunk app (cp_log: orig, src, dst, service_id, action) or CEF CommonSecurityLog; confirm your log exporter populates the gateway address in orig / DeviceAddress. Add known vendor update and licensing destinations to the exclusions and raise unique_destinations if the gateway is noisy. Baseline the gateway for 7-14 days before alerting. No CVE is named in the source reporting, so this is behavioural, not exploit-signature based; pair with vendor-supplied IOC/integrity checks.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- TL-2026-6321 Linux Edge Appliance Web Server Process Spawning Shell With Download or Reverse-Shell Command (locked: unlock with TL Creds)
- TL-2026-6321 DLL Dropped Then Loaded Within 15 Minutes From Same User-Writable Directory (Search-Order Hijack) (locked: unlock with TL Creds)
- TL-2026-6321 Microsoft 365 Mailbox Access From First-Seen IP Shared Across Multiple Users (MURKY PANDA ORB Relay) (locked: unlock with TL Creds)
- TL-2026-6321 Windows Web Server Worker Process Spawning Command Interpreter or LOLBin (locked: unlock with TL Creds)
- TL-2026-6321 Download of [indicator] or vs Binary From Raw-IP Open Directory (locked: unlock with TL Creds)
- TL-2026-6321 [indicator] Execution (Process Name or Characteristic Command-Line Flags) (locked: unlock with TL Creds)
- TL-2026-6321 Unsigned DLL Loaded From Same User-Writable Directory as Host Executable (ShadowPad / KEYPLUG Sideloading) (locked: unlock with TL Creds)
- TL-2026-6321 Single Process Fan-Out to Many Internal Hosts and Service Ports (Fscan-Style Scanning) (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sources

- [Chinese APT Groups Shift From Financial Theft to Cyberespionage Against Global Banking Networks](https://gbhackers.com/global-banking-networks/)
- [Team Cymru: China-Nexus Advanced Persistent Threat Targeting Financial Services](https://www.team-cymru.com/post/china-nexus-advanced-persistent-threat-financial-services)
- [CrowdStrike 2026 Financial Services Threat Landscape Report](https://www.crowdstrike.com/en-us/blog/crowdstrike-2026-financial-services-threat-landscape-report)
- [CrowdStrike 2026 Financial Services Threat Landscape Report: North Korean Adversaries Steal Billions in Digital Assets](https://www.crowdstrike.com/content/crowdstrike-www/locale-sites/us/en-us/press-releases/crowdstrike-2026-financial-services-threat-landscape-report.html)
- [Chinese and Korean threat groups in Asia-Pacific (Aviatrix summary of CrowdStrike report)](https://aviatrix.ai/threat-research-center/chinese-korean-threat-groups-asia-pacific-success-2026)
- [ShadowPad Malware Upgraded to Deliver Ransomware (GBHackers)](https://gbhackers.com/shadowpad-malware-upgraded-to-deliver-ransomware/)
- [Microsoft 365 Token Replay Attacks (GBHackers)](https://gbhackers.com/microsoft-365-token-replay-attacks/)
- [CISA AA20-239A: FASTCash 2.0 North Korea's BeagleBoyz Robbing Banks](https://www.cisa.gov/news-events/cybersecurity-advisories/aa20-239a)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-6321
