FreeSWITCH CVE-2026-116779: one SIP message can run commands, and it is exploited
A flaw in FreeSWITCH lets one crafted message take over a phone server, and attacks are confirmed. What it means for your organisation and what to do first.
Researched and drafted with AI assistance · reviewed by Prince

Developing story — last updated 10 November 2026.
If your organisation runs its phones, contact centre or voice services on FreeSWITCH, treat this as an incident-level priority this week, not routine patching. A single crafted message sent to a FreeSWITCH server can make it run commands of the sender’s choosing. The software’s maintainers say attackers are already doing it, and the US Cybersecurity and Infrastructure Security Agency (CISA) has given federal agencies until 12 November to deal with it. The fix is FreeSWITCH 1.11.3, from SignalWire, the company that maintains the project.
This briefing is for the person who has to decide what happens next. It covers what happened, what it means for the business and what to do, in order of urgency. The technical detail your teams will need is short and comes last.
What happened
FreeSWITCH is open-source software for handling voice calls over the internet (VoIP, voice over internet protocol). It speaks SIP, the Session Initiation Protocol, which phones and telephone carriers use to set up calls. SIP can also carry short text messages, using a request type called MESSAGE, and FreeSWITCH accepts them by default.
On 28 October 2026 SignalWire published a security advisory, GHSA-mg63-h38w-8p9p, describing a flaw in how FreeSWITCH reads the address on those incoming text messages. By writing that address in a particular way, a sender can tell FreeSWITCH to treat the message not as chat but as an administrative command. FreeSWITCH then runs the command with the same permissions as the FreeSWITCH service itself, and the available commands include ones that run any program on the underlying server.
The flaw is tracked as CVE-2026-116779. CVE is the public catalogue that gives each known flaw a single identifier; we opened the official CVE record and the entry in NVD, the US National Vulnerability Database, to confirm it. The record was published on 5 November.
Two facts lift this above a normal patch notice. First, the advisory says plainly: “Exploitation of this path has been observed in the wild.” Second, on 9 November CISA added the flaw to its Known Exploited Vulnerabilities catalogue (KEV), its list of flaws with evidence of real attacks, and set a due date of 12 November, three days later. CISA also marks the entry as requiring forensic triage under its directive BOD 26-04. In plain words, US federal agencies must check affected systems for signs that someone already got in, not just patch them. That directive binds only US federal civilian agencies, but we think the same logic applies to any organisation with an exposed server.
What it means for your organisation
The plain version: someone who can reach your FreeSWITCH server over the network may be able to take control of it, in some set-ups without any password, and attackers are already using the flaw.
According to the advisory, an attacker who succeeds can:
- run any command on the server, with the FreeSWITCH service’s permissions;
- read its configuration and the credentials stored in it;
- take full control of calls, including starting new ones.
In business terms, our reading is this. Whatever passwords and keys sit in the server’s configuration should be treated as exposed, so the damage can reach any account or system those credentials open. An attacker who can start calls from your system can make calls at your expense or in your name. And a server under someone else’s control is a foothold inside your network, from which they can try to reach other systems. These consequences follow from what the advisory describes. No source we read reports a specific victim, loss or fraud case for this flaw.
Does it apply to you?
Your team can answer three questions quickly.
- Do you run FreeSWITCH 1.11.2 or earlier? That includes FreeSWITCH inside a product or service from a supplier, which your team may not know about without asking.
- Can untrusted networks, especially the internet, send SIP traffic to it?
- How is it configured? The text-message feature is switched on in FreeSWITCH’s standard sample configuration, on both its “internal” and “external” profiles. A profile is a set of rules for one group of connections, typically inside the organisation and towards the outside world. The advisory adds that FreeSWITCH’s usual access lists, which restrict who may connect, do not cover these messages.
Configuration also decides whether an attacker needs a password. The advisory says that if an installation still carries the sample “default” user account, which has no password, no account is needed at all. Where messages must be authenticated and no such account exists, the attacker needs one valid account on the server, and any single account will do. So “we require passwords” is not a reason to wait.
One practical point for planning: the 1.11.3 release notes call it an important release with critical security fixes, and say support for Debian 11 (“Bullseye”), a Linux version, has been dropped. If your FreeSWITCH servers run Debian 11, ask your team early whether that changes the upgrade plan.
What to decide and do, in order of urgency
The order below is ours, built from what SignalWire and CISA say. Adjust it to your own estate.
In the next 24 to 48 hours
- Find every FreeSWITCH server. Include test systems, contact-centre platforms and supplier appliances that might carry FreeSWITCH inside. Where you can’t tell, ask the supplier directly and in writing.
- Decide: patch now, or switch the feature off now and patch next. Upgrading to 1.11.3 is the fix. If that can’t happen today, the advisory’s stopgap is to switch off incoming SIP chat on every profile that doesn’t need it, which closes the path on that profile. The cost is that any service relying on SIP text messaging stops working there. For a voice-only deployment that is, in our view, an easy trade, but confirm it with whoever owns the service.
- Preserve evidence before changing exposed servers. Exploitation is confirmed, so assume an internet-facing server may already have been used. CISA’s guidance to federal agencies under BOD 26-04 says not to alter or fix systems before collecting evidence where possible, to collect the most short-lived data, such as memory, first, and warns that patching can destroy traces. A snapshot or memory capture before the upgrade costs little and keeps your options open.
- Limit who can reach the server. If a FreeSWITCH server accepts SIP from the whole internet and doesn’t need to, restrict it at the firewall to your known carriers and sites. This is our general advice, not a vendor workaround, and it doesn’t replace the patch.
This week
- Check for compromise, starting with the servers that were exposed. Ask your team or your incident responder to look for unexpected programs running, new user accounts, changed configuration and unusual outbound connections, and to review call records for calls nobody recognises. No source we read has published indicators of compromise (markers of a known attack, such as network addresses or file fingerprints), so this has to be a check for unusual behaviour.
- Change the credentials held on any exposed server, including SIP account passwords and any carrier or integration credentials in its configuration. If the server may have been read, treat those secrets as known to an attacker.
- Remove sample accounts that have no password. The advisory names the “default” user and an example user for the Skinny phone protocol in the standard sample configuration. Removing them does not fix the flaw, but an attacker would then at least need a real account.
Within 30 days
- Confirm every server runs 1.11.3 or later, and get written confirmation from suppliers that their products are fixed or not affected.
- Decide whether you need incoming SIP chat at all. If not, leave it switched off after patching. Every feature you don’t use is one less way in.
- Ask for a short written report: which servers were exposed, whether any sign of compromise turned up, and which credentials were changed. If the answer is “we found something”, move to full incident response.
- Close the gap that slowed you down. If finding your FreeSWITCH servers took more than a day, the lesson is an inventory of voice systems with a named owner for each. Keep an eye on the vendor advisory and the KEV entry too: this story is still developing.
What is not yet known
- Who is attacking, whom and since when. The advisory says exploitation has been observed, and CISA lists it as exploited, but no source we read names an attacker, a victim, a sector, a start date or a scale. CISA lists ransomware use as unknown.
- How to spot an attack. SignalWire’s advisory includes no detection guidance and no indicators of compromise.
- Whether browser-based calling is affected the same way. The 1.11.3 release notes add the same safeguard to mod_verto, the FreeSWITCH module for browser-based (WebRTC) calls, as well as to the SIP module. The advisory describes only the SIP route, so whether the browser route was exploitable, or exploited, is not stated.
- The account question, in detail. The CVE record and NVD say any authenticated account on a profile is enough. The vendor advisory adds that installations keeping the sample “default” user need no account at all, and the severity score assumes no privileges are needed. These don’t contradict each other, but the official summaries understate the worst case, so plan for the vendor’s version.
- Which products include FreeSWITCH. No source we read lists commercial products or services that embed it, which is why the supplier question above matters.
Technical detail, for your teams
Affected and fixed versions
| Product | Affected | Fixed | Source |
|---|---|---|---|
| SignalWire FreeSWITCH | 1.11.2 and earlier | 1.11.3 | SignalWire advisory; NVD and the CVE record (“before 1.11.3”) |
Severity
The CVSS (Common Vulnerability Scoring System, 0 to 10) score is 9.8, critical, with the vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H. NVD shows this score as supplied by GitHub, which assigned the CVE; NVD lists no separate score of its own. In practice: reachable across a network, easy to carry out, no account and no user action needed, and full loss of confidentiality, integrity and availability. CISA’s own assessment on the CVE record, dated 6 November, records exploitation as active, the attack as automatable and the technical impact as total.
How it works, briefly
According to the advisory, FreeSWITCH’s SIP module splits the user part of an incoming MESSAGE’s To address on a + sign and treats the left half as the name of a chat protocol. The mod_dptools module registers a chat protocol called api that hands the rest to FreeSWITCH’s command interface. So a peer whose MESSAGE is accepted chooses which FreeSWITCH command runs. Where mod_lua is loaded, as it is in the stock configuration, inline Lua scripts offer a second route.
Stopgaps, and what they cost
enable-chat=falseon each profile that doesn’t need incoming SIP chat. The advisory says this fully disables incoming MESSAGE chat on that profile. Cost: SIP text messaging stops on that profile.- Not enough on their own:
auth-messages, which is on by default from 1.10.7, still leaves exploitation open to anyone with an account, even after removing password-less directory users.disable_system_api_commands=trueremoves only thesystemcommand; module loading and call origination stay reachable.
The fix
Version 1.11.3 adds a per-profile setting, enable-chat-api-proto, which decides whether an incoming MESSAGE may select the api protocol. It is off by default and ships commented out in the sample profiles, so leave it that way. Rejected requests receive a 403 response and are logged with the source and the requested command. Messages that name other protocols, or none, are handled as before. NVD links the patch commit and pull request #3135.
Detection
- Version check: anything below 1.11.3 is affected.
- After upgrading: watch for the new 403 log entries. Each one records an attempt, with its source.
- Before upgrading (our suggestion, not vendor guidance): if you log or capture SIP traffic, search for incoming MESSAGE requests whose To user begins with
api+, ignoring case, since the fix itself compares case-insensitively. Absence of hits is not proof of safety if logging was thin. - Indicators: none published in any source we read. If SignalWire or CISA publishes a list, take it from them in its machine-readable form, not from secondary write-ups.
Sources
- GHSA-mg63-h38w-8p9p: Inbound SIP MESSAGE can select the api chat proto and run FreeSWITCH API commands
- CVE-2026-116779
- CVE-2026-116779 (CVE record)
- Known Exploited Vulnerabilities catalog: CVE-2026-116779
- FreeSWITCH v1.11.3 release
- BOD 26-04: Implementation Guidance for Prioritizing Security Updates Based on Risk
Was this useful?