Vulnerability advisory

EEF-CVE-2026-56811

Phoenix transports do not limit channel joins per connection, enabling process-exhaustion denial of service

Published Updated OSV record ↗

Also known as CVE-2026-56811GHSA-6983-jfq8-485w

Severity
  • CVSS 4.0CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Affected versions

Repository

Affected commitsFixed by commit
GIT github.com/phoenixframework/phoenix ↗
14a297e88023cb280a577962a49a0bbdeef9f4eb to before c498ba8cf49f6accbbd0c643a5340b58db891218c498ba8cf49f6accbbd0c643a5340b58db891218
Not specifiedd19ca0a8d9f82c130b7ed339b9f033433e2dea5e
Not specifieda612100cd8a4279091abc1a2ef8fb98a6d01c0a1
Not specified16e295d2fccab185d1292322e2bee5d46c725c8a

Description

Summary

Allocation of Resources Without Limits or Throttling vulnerability in phoenixframework phoenix (Phoenix.Socket module) allows an unauthenticated attacker to cause a denial of service against any endpoint that mounts a Phoenix socket with a reachable channel transport (WebSocket or LongPoll).

This vulnerability is associated with program files lib/phoenix/socket.ex and program routine 'Elixir.Phoenix.Socket':handle_in/4.

Phoenix transports do not limit the number of channels that a single transport process may join. Every phx_join message a client sends over one connection starts a persistent channel process, and the socket process accepts an unbounded number of them. A single unauthenticated client can therefore open one WebSocket or LongPoll connection and stream a large number of phx_join messages, spawning hundreds of thousands of channel processes over that one connection and eventually reaching the BEAM maximum process limit. Once the process table is exhausted the virtual machine can no longer start new processes, denying service to legitimate traffic across the whole node. Because the amplification happens inside a single connection, network-layer connection caps and rate limiting do not mitigate it.

The fix adds a :max_channels_per_transport option (default 100) that bounds the number of channels a single transport process can join, forcing abusive clients to open many connections instead, where external load balancers and reverse proxies can throttle them.

This issue affects phoenix: from 0.11.0 before 1.5.15, from 1.6.0-rc.0 before 1.6.17, from 1.7.0-rc.0 before 1.7.24, and from 1.8.0-rc.0 before 1.8.9.

Workarounds

Front the socket endpoint with a reverse proxy that limits the number and rate of channel-join frames per connection, or disable transports that are not needed (for example remove longpoll: true from the socket declaration where the LongPoll transport is not required). Lowering the BEAM +P maximum process limit does not prevent the exhaustion and can make it easier to trigger.

Configurations

The application must mount a Phoenix socket that defines channels and expose it over HTTP so that a client can reach a channel transport (WebSocket or LongPoll) and send phx_join messages. The WebSocket transport is enabled by default on a mounted socket, while the LongPoll transport is opt-in. Applications that do not mount a socket with channels are not affected.

References

  • Codex
  • Claude Code
  • VS Code
  • GitHub Copilot
  • Cursor
  • Cline
  • Windsurf
  • Google Anti-gravity
  • OpenCode
  • Pi
  • Gemini CLI
  • Hermes Agent
  • Zed
  • Junie
  • Qwen Code
  • Kiro
  • Kilo Code
  • Factory Droid

Set up GitHits for your agents

Run this command:

<> bash
npx githits@latest init

What happens next:

  1. Opens your browser to authorize GitHits
  2. Asks whether to set up GitHits at the user or project level
  3. Detects your supported AI coding tools
  4. Lets you choose which tools to configure
  5. Configures the selected tools