In short
FRP (Fast Reverse Proxy) is an open-source tool, written in Go, that exposes a local server behind NAT or a firewall to the internet. A client called frpc inside the network dials out to a server called frps, which relays traffic back in. Developers use it for testing. Attackers, including Volt Typhoon, have used it for covert access.
Key takeaways
- What it is
- An Apache-2.0 licensed reverse proxy on GitHub with more than 100,000 stars. It carries TCP, UDP, HTTP and HTTPS through an outbound connection.
- How it works
- frpc inside the network connects out to frps on a public server, often on port 7000. Users on the internet then reach internal services through frps.
- Why it matters
- CISA reported that Volt Typhoon planted FRP clients, renamed and packed, on victim servers to open reverse proxies to its own C2.
- What Odin Research saw
- 114 FRP relays in Monograph 12, with 86 exposing the dashboard on port 7500. Most sat on Chinese cloud providers. None is confirmed hostile.
- What to do first
- Allow frpc only where approved, block it elsewhere with application control, and restrict outbound traffic to relay ports such as 7000 and 7500.
What is FRP?
FRP, short for Fast Reverse Proxy, is an open-source tool that exposes a server behind NAT or a firewall to the internet. A client inside the network connects out to a relay server on the internet, and outside users reach the internal service through that relay. It is legitimate software that attackers often abuse.
The project's README describes it as "a fast reverse proxy to help you expose a local server behind a NAT or firewall to the internet". It is written in Go, published under the Apache-2.0 license and maintained on GitHub by fatedier, where it had about 110,700 stars on 21 November 2026.
Developers use FRP to show a test site to a colleague, reach a home lab or publish a service without opening inbound firewall ports. Those same traits make it useful to intruders. FRP turns one outbound connection into a door that can carry Remote Desktop, SMB or a SOCKS proxy back into the network.
How FRP works
FRP has two programs:
- frps, the server, runs on a machine with a public IP address and waits for clients. The README's examples use port 7000 for client connections and port 7500 for a web dashboard.
- frpc, the client, runs inside the private network. It dials out to frps and registers the local services it should expose.
Once the tunnel is up, a connection that reaches frps is passed down the existing outbound session to frpc, which hands it to the internal service. Because the session was opened from inside, a firewall that only blocks inbound traffic does not stop it.
FRP supports TCP, UDP, HTTP and HTTPS proxies. It also offers STCP, which needs a preshared key, XTCP for peer-to-peer transfers, and QUIC and KCP transports. Since version 0.52.0 it reads TOML, YAML and JSON configuration, and the old INI format is deprecated.
Odin Research's Monograph 12 adds two details that matter for defenders. From version 0.50.0, transport TLS is on by default, so the login is only visible in clear text on older builds or when TLS is turned off. The login carries a privilege_key, an MD5 digest of the auth token and a timestamp.
How attackers use FRP
For an intruder, FRP solves a hard problem: how to reach an internal machine without exposing their own server or opening an inbound port. The client sits on a breached host, and the attacker connects through a cheap cloud relay.
The best-documented case comes from CISA's February 2024 advisory AA24-038A on Volt Typhoon. CISA reported that the actors:
- set up FRP clients on victim corporate infrastructure as a command-and-control channel;
- planted the client as
SMSvcService.exeon a Shortel Enterprise Contact Center server and asBrightmetricagent.exeon another server; - packed both files with UPX to hide them;
- used a PRTG network monitoring server as an intermediary for FRP operations, with a
netshPortProxy change to redirect traffic.
Renaming a well-known tool after an ordinary-looking service is a classic living-off-the-land move. The binary still behaves like FRP, which is why behavior-based detection beats file names.
What Odin Research observed
- 114 relays, all on watchlists. Monograph 12 found 114 FRP server relays. Of these, 86 (75.4%) left the web dashboard reachable on port 7500. China hosted 70.2% of them, led by 61 nodes on Alibaba Cloud and 9 on Tencent.
- Owners unknown. Developers also use FRP for test services, and ports and providers alone cannot tell the two apart. None of the 114 relays was confirmed hostile.
- Telling open folders. Four open web directories held FRP-related files. They included a client configuration with an auth token and a cron watchdog that restarts dropped tunnels. Two more were a key dump and a remote code execution playbook, each next to a relay. Whether their owners were attackers or lab administrators is unknown.
- A small baseline. Monograph 06 used three FRP control panels on port 7500 as a baseline next to its larger Ligolo-ng tunnel study.
FRP is one of five dual-use tools in Monograph 12, next to TacticalRMM, Keitaro, Hak5 Cloud C2 and ReverseSSH. Each can stage an attack before any malware lands. Monograph 12 treats the chain as a model, not one observed campaign.
How to detect FRP
- On the endpoint. Alert when
frpcruns, or when a command line referencesfrpc.toml,frpc.iniorserverPort = 7000. This works even when FRP traffic is encrypted. - Renamed binaries. Hunt for UPX-packed executables with service-like names that open long-lived outbound connections, as in the Volt Typhoon case.
- Config files. A file with a server address, an
auth.tokenand asocks5or[[proxies]]entry is an FRP client config. - On the network. Long-lived outbound sessions to ports 7000, 7500 or 17500 from servers, and requests to the dashboard API
/api/proxy/tcp. A plaintext login withprivilege_keyandrun_idshows an older or misconfigured build. - Port proxies. Review
netsh interface portproxyentries on Windows servers. CISA saw them used alongside FRP.
Monograph 12 publishes Suricata signatures for plaintext FRP logins and dashboard API access, a Sigma rule for FRP client execution, a YARA rule for FRP client configs and a Sentinel KQL hunt. They are on the report page and in the Defense Kit.
How to defend against FRP abuse
- Control the binary. Use application control such as WDAC or AppLocker to block unapproved
frpc.exe, even if signed or renamed, and allow it only on named hosts. - Restrict egress. Servers and workstations rarely need outbound access to relay ports 7000, 7500 or 17500. Default-deny outbound traffic from server tiers and log the exceptions.
- Segment the network. Network segmentation limits what a single tunnel can reach, especially RDP and SMB.
- Rotate exposed tokens. If an
frpc.tomlwith a token ever sits in a public place, rotate the token and review the relay. - Find shadow tunnels. Developers' own tunnels are shadow IT. Give them an approved alternative so real misuse stands out.
FRP compared with other tunneling tools
| Tool | How it tunnels | Ports Odin often saw | Notes |
|---|---|---|---|
| FRP | Client dials out to a public relay that forwards traffic back in | 7000, 7500 | 114 relays in Monograph 12; used by Volt Typhoon per CISA |
| Ligolo-ng | Agent opens one encrypted stream to a relay with a virtual network interface | 11601, 443 | 347 recently active relays in Monograph 06 |
| ReverseSSH | Reverse SSH shell on any port | Varies | Two TLS-wrapped listeners confirmed in Monograph 12 |
| Hak5 Cloud C2 | Controller for Hak5 devices and SSH tunnels | 2022 | One server next to GoPhish confirmed in Monograph 12 |
Frequently asked questions about FRP
What is FRP?
FRP, or Fast Reverse Proxy, is an open-source Go tool that exposes a service behind NAT or a firewall to the internet. A client inside the network connects out to a public relay server that forwards traffic back in.
Is FRP malware?
No. FRP is legitimate, Apache-2.0 licensed software with wide developer use. Attackers abuse it as a tunnel, so an unapproved frpc on a server should be investigated.
Is FRP legal to use?
Yes. Using FRP on your own systems is legal. Using it to keep access to a network you are not authorized to use is not.
What is the difference between frps and frpc?
frps is the server that runs on a public machine and accepts connections. frpc is the client that runs inside the private network, dials out to frps and exposes local services through it.
Which ports does FRP use?
The README examples use 7000 for client connections and 7500 for the dashboard. Both are configurable, so attackers can move them.
Which threat actors use FRP?
CISA reported in advisory AA24-038A that Volt Typhoon planted renamed, UPX-packed FRP clients on victim servers. Odin Research also sees FRP relays next to exploit tooling on open servers.
How do I detect FRP?
Look for frpc processes and frpc.toml or frpc.ini files, UPX-packed service binaries with long-lived outbound sessions, and traffic to ports 7000 and 7500. Odin Research publishes Suricata, Sigma, YARA and KQL rules for FRP.
Why do attackers use FRP?
FRP turns one outbound connection from a breached host into a tunnel that can carry RDP, SMB or a SOCKS proxy back into the network. The operator stays behind a cheap relay, and inbound firewall rules do not stop it.
Sources
- fatedier, frp repository and README, GitHub. github.com/fatedier/frp (checked 21 November 2026)
- CISA, Advisory AA24-038A on PRC state-sponsored actors (Volt Typhoon), 7 February 2024. www.cisa.gov/news-events/cybersecurity-advisories/aa24-038a (checked 21 November 2026)
Related tools and terms
What is noVNC?
noVNC is an open-source VNC client that runs in a web browser. Phishing kits abuse it to show victims a real browser hosted by the attacker.
Tunneling toolWhat is Ligolo-ng?
Ligolo-ng is an open-source tunneling tool that builds a network tunnel from a compromised host through a TUN interface.
Tunneling toolWhat is ReverseSSH?
ReverseSSH is an open-source tool that gives an operator an SSH shell through a connection the target opens outward.
Shell managerWhat is Platypus?
Platypus is an open-source Go tool for managing many reverse-shell sessions at once.
Device controllerWhat is Hak5 Cloud C2?
Hak5 Cloud C2 is a controller for Hak5 implant devices, such as drop boxes, that call home over the internet.
Remote managementWhat is TacticalRMM?
TacticalRMM is an open-source remote monitoring and management (RMM) suite built on MeshCentral. MSPs use it, and attackers abuse it for signed, persistent access.
Odin Research
Get the rules, not just the definition
Every Odin Research report ships with verified indicators and detection rules you can load today.