/

CVE-2026-65644

Severity:

HIGH RISK

(7.5)

Stored XSS (CWE-79) in Rocket.Chat Livechat Queue UI

CVE-2026-65644

Analyzed

Vulnerability type

Category:

Stored Cross-Site Scripting

CWE:

CWE-79

Impact:

Client-Side HTML Injection, Phishing, and Potential Session or Data Exposure

Severity:

HIGH RISK

(7.5)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Rocket.Chat contains a stored cross-site scripting vulnerability in the Livechat visitor API, where an unauthenticated name value is saved without adequate sanitization and later rendered as HTML.

An attacker may use this behavior to place a deceptive, clickable link and social-engineering content in the Omnichannel Queue visible to agents. The risk arises when an agent views a queue containing the attacker-controlled visitor entry.

Affected context

The issue occurs in Rocket.Chat deployments where the Livechat visitor creation endpoint accepts visitor names without authentication and stores them in a form that is later rendered through an HTML-capable client-side component. It is particularly relevant to Omnichannel environments where agents view visitor names in the queue and where browser security controls do not fully prevent interaction with attacker-controlled links or markup.

Exploit preconditions

  • Attack surface: The unauthenticated Livechat visitor REST API
  • Required access: Ability to submit a visitor name to the endpoint; no agent credentials are required to create the entry
  • Trigger point: An agent views the affected visitor entry in the Omnichannel Queue side panel
  • Why exploitation is possible: The submitted name is stored without sufficient sanitization and later inserted into the browser DOM as HTML rather than being rendered strictly as text

Why this happens

The underlying flaw is insufficient output encoding and sanitization of user-controlled data before rendering it in an HTML-capable frontend component. A visitor name is treated as trusted markup and passed to a rendering path using dangerouslySetInnerHTML, allowing attacker-controlled HTML content to become part of an agent's browser DOM.

This maps to CWE-79: Improper Neutralization of Input During Web Page Generation, commonly known as cross-site scripting. The stored nature of the data means the malicious content persists until it is displayed to users, rather than requiring the attacker to target each viewer directly.

Potential impact

  • An attacker may display deceptive links or social-engineering messages to agents.
  • Agents could be redirected to attacker-controlled domains and exposed to credential-phishing attempts.
  • Malicious content could abuse the privileges and trust associated with an agent's browser session.
  • The vulnerability could contribute to unauthorized actions, data exposure, or account compromise depending on browser protections and application behavior.
  • Security monitoring and support workflows could be disrupted by misleading queue entries.

General mitigation guidance

Mitigation generally focuses on treating visitor-controlled fields as plain text and reducing trust in data that crosses from the API into the browser.

Common defensive patterns include:

  • Rendering visitor names through framework-safe text interpolation instead of raw HTML insertion.
  • Applying context-appropriate output encoding and allowlist-based sanitization where limited markup is genuinely required.
  • Validating and normalizing URL-bearing values, with restrictive link policies and safe handling of external destinations.
  • Enforcing a strong Content Security Policy and related browser protections as defense in depth.
  • Limiting the privileges available to agent sessions and separating sensitive administrative actions from ordinary queue interactions.
  • Applying consistent sanitization and encoding at every display context rather than relying only on API-side validation.

How teams detect issues like this

Organizations can look for this class of issue through:

  • Static analysis rules identifying dangerouslySetInnerHTML, raw DOM insertion, and template rendering of request or database values.
  • Code review checks that trace visitor-controlled fields from API handlers through storage and frontend rendering.
  • Tests that submit markup-like and URL-bearing values, then verify that the interface displays them as inert text.
  • Runtime monitoring for unusual visitor creation activity, suspicious names, external-domain links, or repeated submissions.
  • Frontend security policies that flag unsanitized HTML rendering and enforce approved sanitization libraries.

Automated application security and code review tools may help identify raw HTML sinks, but manual data-flow review remains important for stored content.

Sources

SourceLink
github.comgithub.com/RocketChat/Rocket.Chat/pull/41595
hackerone.comhackerone.com/reports/3872858

Frequently asked

questions

Yes. The visitor creation endpoint accepts unauthenticated input over the application interface, while exploitation of the stored content generally occurs when an agent later views the affected queue entry.

Not to submit the malicious visitor name. The attacker does, however, rely on an agent or another privileged user viewing the stored entry for the content to have an effect.

Treat visitor-controlled names as text and avoid raw HTML rendering. If a business requirement permits limited markup, use strict allowlist sanitization, context-aware encoding, and safe URL handling.

Map untrusted data through its full lifecycle—from API input and storage to every frontend output sink—and enforce secure rendering policies in code review, automated testing, and application security tooling.

Find out what's already
exploitable in your codebase.