CVE-2026-114713: UnrealIRCd IRCv3 Labeled-Response Server-to-Server Command Injection
On November 13, 2026, CISA cataloged CVE-2026-114713 in the Known Exploited Vulnerabilities (KEV) database with an urgent due date of November 16, 2026. The vulnerability affects UnrealIRCd versions 6.0.0 through 6.2.7. Due to unescaped delimiters within the IRCv3 labeled-response module, an ordinary connected user on any multiserver IRC network can inject protocol line-breaks (CRLF) and delimiters into the message 'label' tag. When the local daemon relays the tag across server links, the remote routing hubs parse the injected stream as native server-to-server (S2S) protocol frames executing with full server and IRC Operator (IRCOp) authority. Threat actors actively exploit this flaw to execute mass network-wide disconnections (KILL), deploy permanent IP bans (GLINE), desynchronize routing hub links, and hijack administrative channels. Organizations and community infrastructure operators must immediately deploy UnrealIRCd 6.2.7.1, apply in-memory hot-patch bundle-20261107, or enforce an emergency spamfilter regex.
HERMES THREAT SCORE & NETWORK INFRASTRUCTURE RISK
Target:UnrealIRCd labeled-response Module and IRCv3 Tag Serialization Engine While CVSS v3.1 rates this flaw at 8.6 (High) based on a requirement for user authentication, Hermes assigns an HTS score of 93 (CRITICAL). In production IRC deployments, users can register and connect anonymously in seconds without vetting. Once connected, an unprivileged user bypasses all role boundaries, executing commands with root server-to-server authority across all linked network nodes, causing catastrophic disruption across communication backbones.
1. Executive Summary & Operational Context
Section titled โ1. Executive Summary & Operational ContextโUnrealIRCd is one of the most widely deployed open-source Internet Relay Chat daemons in existence, powering critical developer communication channels, security incident response networks, and distributed messaging grids.
On November 13, 2026, CISA cataloged CVE-2026-114713 following telemetry confirming active in-the-wild exploitation. Termed by the core UnrealIRCd development team as one of the most severe protocol flaws disclosed in 15 years, the issue allows an unprivileged client on a multi-server IRC network to break out of client boundaries and inject arbitrary server-to-server (S2S) commands.
The flaw is located in the labeled-response module (src/modules/labeled-response.c), which implements the IRCv3 client capability for correlating requests with asynchronous server responses via @label message tags. When echoing client-supplied label tags or forwarding them across inter-server sync links, the module fails to strip or escape carriage return (\r), newline (\n), space, or delimiter characters. When a neighboring linked hub receives the message line, the unescaped delimiter causes the protocol parser to interpret trailing payload bytes as an authentic, high-privilege server command originating from the peer serverโs UID.
2. Technical Vulnerability Analysis & Root Cause
Section titled โ2. Technical Vulnerability Analysis & Root CauseโUnder the IRCv3 specification, message tags are prepended to protocol lines using the @ prefix and separated by semicolons (e.g. @label=req123;time=...).
In vulnerable UnrealIRCd versions (6.0.0 through 6.2.7), the labeled-response module copies client-provided label strings directly into outgoing inter-server response buffers without validating character boundaries:
// Vulnerable tag echo sequence in src/modules/labeled-response.cstatic void send_labeled_response(Client *to, const char *label, const char *format, ...){ char tag_buf[512]; // Flaw: label string is inserted without sanitizing CRLF (\r\n) or spaces snprintf(tag_buf, sizeof(tag_buf), "@label=%s ", label);
char msg_buf[1024]; va_list ap; va_start(ap, format); vsnprintf(msg_buf, sizeof(msg_buf), format, ap); va_end(ap);
// Forwarded across inter-server links to neighboring IRCd hubs: sendto_server(to, "%s%s", tag_buf, msg_buf);}An attacker connects to Server A and submits a command containing injected newline bytes within the label tag:
@label=test\r\n:ServerA GLINE *@badactor.ip 86400 :Network Hijack\r\n PING 12345When Server A formats the inter-server sync packet and forwards it to Server B, Server Bโs parser receives raw CRLF delimiters. Because Server B trusts Server A implicitly across the authenticated S2S link, Server B processes the second line (GLINE ...) as an authentic administrative directive emitted by Server A, enforcing global bans, executing KILL commands against arbitrary users, or forcing clients into rogue channels (SAJOIN).
3. Attack Execution Flow & Weaponization Mechanics
Section titled โ3. Attack Execution Flow & Weaponization Mechanicsโflowchart TD A["Unprivileged User<br/>(Connected to Server A)"] -->|"Step 1: Sends IRCv3 Message with Crafted Tag<br/>@label=xyz\\r\\n:ServerA KILL TargetNick :Compromised"| B["UnrealIRCd Server A"] B -->|"Step 2: labeled-response Module Echoes Tag<br/>Fails to Escape Delimiters"| C["S2S Hub Transport Link"] C -->|"Step 3: Transmits Raw CRLF Stream"| D["UnrealIRCd Server B (Linked Hub)"] D -->|"Step 4: Parser Interprets Injected Line<br/>as Legitimate S2S Directive"| E["Executes with Server / IRCOp Authority"] E -->|"Step 5a: Issues Global GLINE / KILL"| F["Mass User Disconnections"] E -->|"Step 5b: Injects Rogue Operator Privileges"| G["Network Administrative Hijacking"] E -->|"Step 5c: Desynchronizes Routing Mesh"| H["Network Split & Denial of Service"]4. Forensic Signatures, Detection Queries & IOCs
Section titled โ4. Forensic Signatures, Detection Queries & IOCsโ# UnrealIRCd Live Spamfilter Command (IRCOp authority required)/SPAMFILTER add -regex T block - Invalid_label ^label=(?![^\r\n;\x20]*\z)title: UnrealIRCd Delimiter Injection Command Abuseid: f1147130-1147-4140-8114-114713000001status: experimentaldescription: Detects unusual server-to-server IRC commands executed without operator session authentication in IRC daemon logslogsource: category: application product: unrealircddetection: selection_commands: Message|contains: - 'GLINE' - 'KILL' - 'SAJOIN' - 'SVSNICK' selection_delimiters: Message|contains: - 'Invalid_label' - '@label=' - '%0D%0A' condition: selection_commands and selection_delimiterslevel: critical// Detect anomalous IRC server log events indicating label injection and unauthenticated GLINE/KILL burstsSyslog| where ProcessName =~ "unrealircd" or SyslogMessage has "ircd"| where SyslogMessage has "Invalid_label" or (SyslogMessage has_any ("KILL", "GLINE", "SVSNICK") and SyslogMessage has "@label=")| project TimeGenerated, HostName, ProcessName, SyslogMessage# Grep for spamfilter matches and tag anomalies in IRCd logsgrep -iE "invalid_label|@label=" /var/log/unrealircd/ircd.log
# Check active server-to-server links and recent burst disconnectsgrep -i "KILL" /var/log/unrealircd/ircd.log | tail -n 50
# Verify running UnrealIRCd version/path/to/unrealircd versionThreat Intelligence & Indicators of Compromise (IOCs)
Section titled โThreat Intelligence & Indicators of Compromise (IOCs)โ| Type | Indicator | Context |
|---|---|---|
| IP Address | 91.240.118[.]172 | In-the-wild IRC S2S exploit source IP |
| IP Address | 185.220.101[.]5 | Tor exit node generating tag injection packets |
| IP Address | 194.26.29[.]112 | Scanning node probing for multiserver UnrealIRCd topologies |
| Pattern | ^label=(?![^\r\n;\x20]*\z) | Malicious label tag regular expression pattern |
5. Comprehensive Remediation, Hardening & Mitigation Plan
Section titled โ5. Comprehensive Remediation, Hardening & Mitigation Planโ-
Apply Hot-Patch (Zero Downtime / In-Memory): For running *NIX-based UnrealIRCd 6 instances, deploy the hot-patch bundle immediately without restarting the daemon:
Terminal window ./unrealircd hot-patch bundle-20261107 -
Upgrade to UnrealIRCd 6.2.7.1: Download and compile the official release UnrealIRCd 6.2.7.1 (or later), which incorporates strict tag character whitelisting and eliminates CRLF propagation across inter-server links.
-
Deploy Immediate Spamfilter Workaround: If patching cannot be executed immediately, apply the vendor-recommended regex spamfilter across all network servers:
/SPAMFILTER add -regex T block - Invalid_label ^label=(?![^\r\n;\x20]*\z) -
Audit Inter-Server Link Configurations: Review
link { ... }configuration blocks inunrealircd.conf, ensuring all server links mandate TLS (options { ssl; }) and strong cryptographic preshared passwords (password "..." { type argon2; };).