Joplin CVE-2026-105784

Exploitation requires the Joplin desktop app and a note whose content includes a jsoncanvas fence rendered by the WhiteboardEditor; the malicious Markdown must reach the victim (e.g., via a shared notebook, synced data, or imported note) and the victim must actively select/open that note (UI:R) for the TextNode/FileNode components to render the card via dangerouslySetInnerHTML. Limiting factors that reduce real-world risk: Content Security Policy blocks inline script, capping impact at CSS injection, UI redressing, open-signaling, and possible disclosure of exposed attribute values - not code execution; the attacker cannot self-trigger without the user opening the note; and only versions prior to 3.7.13 are affected.

This is a genuinely low-to-moderate risk issue that should not be over-prioritized. The vendor CVSS of 4.6 (CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:U/C:L/I:L/A:N) correctly reflects a bounded-impact flaw: it requires the victim to select/open an attacker-supplied note (UI:R) and yields only low confidentiality and integrity impact with no availability impact. Crucially, the advisory and description explicitly state that Content Security Policy blocks inline script execution, so the realistic ceiling is CSS injection and UI redressing - NOT remote code execution, despite the 'RCE' tag in the input, which is contradicted by the primary description and should be disregarded. No EPSS score, CISA KEV listing, or public POC was provided in the input, so there is no public exploit identified at time of analysis and no evidence of active exploitation; this is a priority driven by patch hygiene, not by threat intelligence. One signal conflict to note: the source-code evidence (the stripUnsafeStyles patch and its tests targeting <style>/@import) fully corroborates the described CSS-injection mechanism, which is reassuring, but the attack is social-engineering-gated (victim must open a crafted note), making it targeted rather than wormable.

An attacker crafts a Joplin note containing a jsoncanvas fence whose whiteboard card holds malicious Markdown that renders to a <style> element with a remote @import. When the victim selects the note, the WhiteboardEditor renders the card into the main application document via dangerouslySetInnerHTML, injecting CSS that restyles trusted app chrome, beacons the attacker when the note is opened, and may leak exposed attribute values. No public exploit was identified at time of analysis, and CSP prevents this from escalating to code execution.