What is UnrealIRCd?
UnrealIRCd is an open-source IRC server daemon and, by its own description and by most third-party counts, the most widely deployed one. It powers a large share of the public IRC networks still running, along with private networks used by open-source projects, gaming communities, and, less happily, botnet command-and-control channels. It is also commonly packaged by Linux distributions, so instances turn up on hosts whose owners did not deliberately choose UnrealIRCd.
What makes an IRC daemon a valuable target is the trust model between servers. An IRC network is a mesh of linked servers that exchange a privileged server-to-server protocol. Those peers accept each other's commands as authoritative: a server can disconnect users, change modes, reveal hidden host information, and manipulate channels without any further authorization check. A client-side bug that reaches that channel does not stay client-side.
Overview
CVE-2026-114713 is a delimiter injection flaw in UnrealIRCd's labeled-response module. The module implements the IRCv3 labeled-response capability, in which a client attaches a label message tag to a command and the server echoes that label back on the replies so the client can match answers to requests. UnrealIRCd did not escape the label value before echoing it, so a crafted label can carry protocol delimiters and append extra lines to the reply.
On a network of two or more linked servers, those injected lines propagate across the server-to-server link, and the receiving peer parses them as commands issued by a server rather than by a user. Any ordinary, unprivileged user on the network can therefore execute commands with server or IRC operator authority, including KILL. The UnrealIRCd project rates the issue critical and calls it the most critical issue found in UnrealIRCd in the past 15 years. NVD scores it 8.6 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:N/I:H/A:N), with the scope change reflecting that the compromised component is the client connection but the affected component is the peer server.
Sources disagree on the weakness classification. NVD and CISA record CWE-140 (improper neutralization of delimiters), Red Hat's entry lists CWE-93 (CRLF injection), and VulDB lists CWE-74 with a lower score of 7.3. All three describe the same defect.
Affected Versions
| Product | Vulnerable | Fixed |
|---|---|---|
| UnrealIRCd 6 | 6.0.0 through 6.2.7 | 6.2.7.1, or hot-patch bundle-20261107 |
| UnrealIRCd 5 | All releases, "similarly affected" per the project | No fix; end of life since 2023-07-01, migrate to 6.2.7.1 |
| UnrealIRCd 3.2.x and 4 | Not assessed by the project | Unmaintained, out of scope |
| Alpine Linux package | Earlier than 6.1.10-r1 | 6.1.10-r1 |
Technical Details
The root cause is CWE-140: the code copies attacker-supplied text into a structured output stream without neutralizing the characters that stream uses as separators. The IRC wire protocol is line-oriented, with CR and LF terminating a message and spaces separating parameters, so a label value containing those bytes is not data any more, it is additional protocol.
The attack needs no authentication and no user interaction. An attacker connects as a normal client, negotiates the labeled-response capability, and sends a single command whose label tag contains the payload. The server builds its reply, embeds the unescaped label, and the forged trailing lines ride outbound to the linked peer, where they arrive on a trusted link and are executed. Attack complexity is low and the whole thing is one request. The only structural precondition is that the network has at least two linked servers, which is the normal configuration for any network large enough to matter; a single standalone server has no peer to inject into.
Discovery
The UnrealIRCd project found the bug itself, during an internal code audit that it says was carried out with the help of AI. No external researcher or reporting organization is credited, and there is no third-party research write-up or published proof-of-concept. A Sploitus entry nominally labeled a PoC for this CVE contains only the CVE description text and is not exploit code.
The same 6.2.7.1 release closed eight further issues from that audit: two IP address leaks, one of which lets a normal user decloak another user's real address, crashes reachable through CHGIDENT and whowasdb, a DNS blacklist recheck that could hang the server, a connection-counting bug that could lock out new clients, a cap on bans set by services, and a TLS renegotiation flood fix. None of those carry separate public CVE IDs.
Exploitation Context
At release on 2026-11-07 the project reported no known exploitation and no public proof-of-concept. That changed within days: the project updated its advisory to state that it was seeing attackers exploit the issue on the internet at that moment, and CISA added the CVE to the KEV catalog on 2026-11-13 with an unusually short three-day remediation deadline.
Public detail is thin. There is no named threat actor, no victim count, no internet-wide exposure figure, and no published indicators of compromise. EPSS put the 30-day exploitation probability at roughly 0.3% before the KEV listing, which illustrates how poorly scoring models track a bug whose exploitation depends on an attacker simply joining a public network. CISA does not flag known ransomware use. Note that some vulnerability trackers still showed "no evidence of exploitation" because they were last refreshed on 2026-11-08, before the project's update.
The practical risk picture is that exploitation requires nothing more than a client connection to a network running an unpatched server. Public IRC networks accept such connections by definition, which makes the population of exposed, reachable targets effectively the whole installed base.
Remediation
- Upgrade to UnrealIRCd 6.2.7.1, or apply hot-patch bundle-20261107, which covers 6.0.0 through 6.2.7 and takes effect without restarting the daemon on almost all UnrealIRCd 6 networks.
- Apply the project's spamfilter workaround first, before patching, if you cannot patch immediately. An IRC operator adds the rule published in the advisory to block invalid labels; on 6.0.4 and newer it protects every linked server on the network, not just the one it is entered on. The project recommends this workaround-first sequence even when you intend to hot-patch straight away.
- On UnrealIRCd 5, blacklist the labeled-response module and restart the server as an interim measure, then migrate to 6.2.7.1. Version 5 has been end of life since 2023-07-01 and will not receive a fix.
- If patching is not possible at all, unlinking servers so the network runs standalone removes the precondition for the injection, at the cost of splitting the network. Treat this as a last resort, not a fix.
- Review server logs for unexpected
KILLcommands, mode changes, or operator-level actions that no operator performed, and for client connections negotiating labeled-response with unusual label values. Preserve those logs before remediating, in line with CISA's Forensics Triage Requirements under BOD 26-04. No indicators of compromise have been published, so this is best-effort hunting. - Confirm the running version afterwards rather than trusting the package manager, since the hot-patch and a full upgrade report differently, and verify that every server on the network is patched. One unpatched peer leaves the link exploitable.
- Federal civilian agencies must remediate by 2026-11-16. Given a hot-patch that needs no restart and active exploitation, there is no reason to use the full window.
Key Details
| Property | Value |
|---|---|
| CVE ID | CVE-2026-114713 |
| Vendor / Product | UnrealIRCd - UnrealIRCd |
| NVD Published | 2026-11-08 |
| NVD Last Modified | 2026-11-13 |
| CVSS 3.1 Score | 8.6 |
| CVSS 3.1 Vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:N/I:H/A:N |
| Severity | HIGH |
| CWE | CWE-140 find similar ↗ |
| CISA KEV Added | 2026-11-13 |
| CISA KEV Deadline | 2026-11-16 |
| Known Ransomware Use | No |
CVSS 3.1 Breakdown
Required Action
Timeline
| Date | Event |
|---|---|
| 2026-11-04 | UnrealIRCd announces an upcoming security release for 2026-11-07, without disclosing details |
| 2026-11-07 | UnrealIRCd 6.2.7.1 and hot-patch bundle-20261107 released, fixing nine security issues |
| 2026-11-08 | CVE-2026-114713 published with a CVSS 3.1 base score of 8.6 |
| 2026-11-13 | Added to the CISA Known Exploited Vulnerabilities catalog after the project reports active exploitation |
| 2026-11-16 | CISA BOD 26-04 remediation deadline |
References
| Resource | Type |
|---|---|
| NVD - CVE-2026-114713 | Vulnerability Database |
| CISA KEV Catalog Entry | US Government |
| UnrealIRCd 6.2.7.1 released and hot-patch to fix 9 security issues, including one CRITICAL (CVE-2026-114713) | Vendor Advisory |
| UnrealIRCd - Project News and Downloads | Vendor Advisory |
| SecurityOnline - UnrealIRCd Vulnerability CVE-2026-114713 Exploited in the Wild, Lets Users Run IRCOp Commands | News |
| Brinztech Alert - UnrealIRCd Releases 6.2.7.1 and Emergency Hot-Patch Addressing Nine Flaws | News |
| OSV - ALPINE-CVE-2026-114713 | Vulnerability Database |
| CISA Alert - CISA Adds Two Known Exploited Vulnerabilities to Catalog | US Government |