In short
VShell is a Chinese-language remote access tool (RAT) that lets an operator run commands and move files on Linux, Windows and macOS machines. Its agents often talk to the controller over WebSocket, which blends with normal web traffic. Sysdig and NVISO have tied it to espionage campaigns, including activity by the actor UNC5174.
Key takeaways
- What it is
- A remote access tool with stagers for Linux, Windows and macOS. Sysdig traced its public release to a GitHub user named veo, who later deleted it.
- How it works
- A small stager pulls the full agent, which can run from memory on Linux. The agent then holds a WebSocket or TCP session to the controller.
- Why it matters
- Sysdig reported VShell in a 2025 UNC5174 campaign. NVISO says several state-aligned and independent actors now use it.
- What Odin Research saw
- 769 suspected or candidate VShell endpoints in Monograph 14, none confirmed. Most listen on port 9999, and one open server logged downloads of a Windows agent.
- What to do first
- Alert on WebSocket upgrades to high ports from servers, and on VShell stager file names. Do not block on port 9999 alone.
What is VShell?
VShell is a Chinese-language remote access tool that lets an operator run commands, upload and download files and keep access on Linux, Windows and macOS machines. Agents often reach the controller over WebSocket, so the traffic looks like ordinary web app updates. Espionage and criminal actors both use it.
Sysdig describes VShell as a remote access trojan that is popular among Chinese-speaking cybercriminals on several forums. NVISO calls it a Chinese-language intrusion tool that has been seen more and more over the past year, mostly for long-term espionage.
VShell started life as a public project. According to Sysdig, it was published on GitHub by a user named veo. The developer later deleted the release, citing legal reasons, but copies kept circulating. That is a common path for dual-use tools: once the code is out, the author no longer controls who runs it.
How VShell works
VShell follows the usual stager-and-agent design.
- Stager. A small first-stage program runs on the target and asks the controller for the full agent. NVISO's network rules cover stagers for Windows (32- and 64-bit), Linux (amd64, i386, arm64 and arm) and macOS (amd64 and arm64). Each stager request starts with a short platform marker.
- Agent in memory. In the campaign Sysdig studied, the VShell payload never touched disk. It was loaded through the Linux
memfd_createcall and disguised as a kernel worker process named[kworker/0:2]. - Control channel. The agent keeps a session with the controller. Sysdig saw WebSocket command and control over encrypted HTTPS on port 8443. Odin Research's monographs list paths such as
/ws_agentand/ws_clientfor the WebSocket agent endpoint.
Once connected, the operator can run shell commands, move files and use the host as a pivot. Because a WebSocket upgrade is a normal HTTP feature, a firewall that only checks ports and protocols sees nothing odd. That maps to MITRE ATT&CK's Application Layer Protocol (T1071).
Who uses VShell: UNC5174 and others
In April 2025, the Sysdig Threat Research Team described a campaign it identified in late January 2025. It attributed the activity to UNC5174, which it calls a Chinese state-sponsored actor. Citing Mandiant and HivePro, Sysdig said the group is believed to be a contractor working for the Chinese government.
In that campaign, a dropper called SNOWLIGHT delivered VShell as a fileless, memory-only payload. Sysdig also found an older VShell sample from November 2024 that used the WebSocket protocol.
In November 2025, NVISO published its own analysis. It described UNC5174 as a suspected initial access broker linked to China's Ministry of State Security. NVISO also stressed that VShell use cannot be tied only to UNC5174, because multiple state-aligned and independent actors deploy it. It reported activity concentrated in South America, Africa and the Asia-Pacific region.
For defenders, the practical point is simple. Finding VShell does not tell you who is behind it. It does tell you someone wants hands-on, long-term access.
What Odin Research observed
VShell appears in five Odin Research studies. The data shows a large listener fleet, very little of it confirmed, and a few telling staging servers.
- 769 endpoints, none confirmed. Monograph 14 counted 769 VShell endpoints on 685 addresses: 282 suspected and 487 candidates. The top ports were 9999 (227), 8084 (71) and 9097 (68). Most sat in Hong Kong (270), China (218) and the United States (187).
- A staging server with a download log. In the same study, one open web server held 719 files, including maps of hospital web portals, a Windows VShell agent and two small Linux stagers. Its own access log showed two outside hosts downloading the Windows agent. The log shows the file was served, not that it ran.
- One TLS stack, one provider. Monograph 02 found 276 VShell records on 270 hosts. One JARM covered 233 of them, and one hosting provider held 220. A Go TLS listener on port 9999 accounted for 219.
- Scripted Linux deployment. Monograph 03 described a staging server whose deploy script tests which folders allow execution, then fetches a VShell binary of more than 7 MB. The scripts show intent and tooling, not activity on a target.
- Next to callback servers. Monograph 17 found VShell WebSocket listeners on the same addresses as DNSLog and Interactsh callback servers. Monograph 19 saw a VShell listener in the same Alibaba Cloud range as a Hashtopolis candidate. Sharing a cloud range does not show one operator.
The lesson: VShell listeners are good hunting leads. Treat them as leads until a second signal, such as a stager download or a known controller, confirms them.
How to detect VShell
- WebSocket upgrades from servers. Alert on outbound HTTP upgrade requests to ports 9999, 8084 or 9097, and to paths such as
/ws_agentor/ws_client. Allowlist your own WebSocket services first. - Stager downloads. Look for
GETrequests forvs_agent.exe,vs_linux_amd64orst_5_linux_amd64, and for URIs containingv_ver=. - Fake kernel threads. On Linux, a process named like
[kworker/...]that has a parent in user space, an open network socket or an executable mapped frommemfdis not a real kernel worker. - Odd execution folders. Watch for new executables in
/dev/shm,/tmpor hidden folders under a home directory, especially right after a web server process spawns a shell.
Monograph 14 publishes 12 Suricata signatures for VShell WebSocket and stager traffic, a Sigma rule for VShell stagers on Linux and a YARA rule for VShell delivery logs. Monograph 02 adds a Sigma rule for Windows living-off-the-land binaries reaching the VShell port 9999. NVISO has also published its own network signatures. Odin's rules are in the Defense Kit.
How to defend against VShell
- Patch internet-facing servers fast. VShell arrives after an attacker already has a foothold, often on an internet-facing server. See how to respond to actively exploited vulnerabilities.
- Limit server egress. Web, database and middleware servers rarely need to open WebSocket sessions to unknown cloud hosts on high ports. Default-deny outbound traffic from those tiers.
- Harden Linux temp folders. Mount
/tmpand/dev/shmwithnoexecwhere you can, and log process starts on Linux servers. - Hunt with context. Feed suspected VShell listeners into threat hunting rather than blocklists, and confirm with endpoint evidence before you isolate a host.
VShell compared with similar tools
| Tool | Main channel | Ports Odin often saw | Notes |
|---|---|---|---|
| VShell | WebSocket and TCP | 9999, 8084, 9097 | Memory-only Linux agent in the Sysdig case; none of 769 endpoints confirmed in Monograph 14 |
| SuperShell | Browser console for reverse SSH shells | 443, 8888 | 171 of 180 endpoints in Hong Kong in Monograph 22 |
| Sliver | mTLS in its default implant profile | 31337, 8888 | Team C2 framework; TLS 1.3 hides its certificate |
| Cobalt Strike | Beacon over HTTPS and other profiles | 50050 and many high ports | 28 confirmed servers share one certificate template in Monograph 19 |
Frequently asked questions about VShell
What is VShell?
VShell is a Chinese-language remote access tool that lets an operator run commands and move files on Linux, Windows and macOS machines, usually over a WebSocket or TCP channel to a controller.
Is VShell malware?
It was released publicly as a tool, but it is used as malware. Sysdig and NVISO have both documented its use in espionage intrusions, so a VShell agent on your network should be treated as hostile.
Who made VShell?
Sysdig traced the public release to a GitHub user named veo, who later deleted it for legal reasons. Copies continue to circulate.
Is VShell linked to China?
Sysdig tied a 2025 VShell campaign to UNC5174, which it describes as a Chinese state-sponsored actor. NVISO cautions that several state-aligned and independent actors use VShell, so the tool alone does not identify the operator.
What is SNOWLIGHT?
SNOWLIGHT is the dropper Sysdig observed delivering VShell in the UNC5174 campaign. It loaded VShell as a memory-only payload on Linux.
Which ports does VShell use?
There is no single fixed port. Sysdig saw port 8443. In Odin Research data, most VShell listeners sat on 9999, 8084 and 9097. Ports alone are not enough to confirm VShell.
How do I detect VShell?
Alert on WebSocket upgrades from servers to high ports, stager downloads such as vs_agent.exe or vs_linux_amd64, and fake kworker processes with network sockets. Odin Research publishes Suricata, Sigma and YARA rules for these signals.
VShell vs Cobalt Strike: what is the difference?
VShell is a lighter remote access tool with stagers for Linux, Windows and macOS and a WebSocket channel. Cobalt Strike is a larger C2 framework built around its Beacon implant. Intruders use both after they gain initial access.
Sources
- Sysdig Threat Research Team, "UNC5174's evolution in China's ongoing cyber warfare: From SNOWLIGHT to VShell", 15 April 2025. www.sysdig.com/blog/unc5174-chinese-threat-actor-vshell (checked 21 November 2026)
- NVISO, "NVISO analyzes VShell post-exploitation tool", 5 November 2025. www.nviso.eu/blog/nviso-analyzes-vshell-post-exploitation-tool (checked 21 November 2026)
Related tools and terms
What is Sliver?
Sliver is an open-source command-and-control (C2) framework from Bishop Fox, written in Go. Red teams use it for adversary emulation, and criminal and state operators use it as a free alternative to Cobalt Strike.
C2 frameworkWhat is Brute Ratel C4?
Brute Ratel C4 is a commercial red-team C2 framework built to evade endpoint detection. Leaked copies have been used by criminal operators.
C2 frameworkWhat is Havoc?
Havoc is an open-source post-exploitation C2 framework with a Windows implant called Demon.
C2 frameworkWhat is Mythic?
Mythic is an open-source, multi-user C2 platform with pluggable agents and transports, built for red-team operations.
Offensive frameworkWhat is Metasploit?
Metasploit is an open-source penetration testing framework maintained by Rapid7. It bundles exploits, payloads and the Meterpreter post-exploitation agent.
Offensive frameworkWhat is Viper?
Viper is an open-source browser front end for Metasploit. It adds a multi-operator web interface and a module runner on top of the Metasploit RPC service.
Odin Research
Get the rules, not just the definition
Every Odin Research report ships with verified indicators and detection rules you can load today.