imgproxy CVE-2026-105708

Exploitation requires that imgproxy be used to process and serve attacker-influenced SVG content whose output is rendered in a browser context, so that a flawed sanitizeElement in the SVG handler lets active content (script/event handlers) survive sanitization. Key limiting factors: a victim must actually render the processed SVG in a browser (CVSS UI:P - passive user interaction), and the deployment must pass SVG through imgproxy's SVG handling path (SVG processing/sanitization in use) rather than blocking or rasterizing SVGs; impact is limited to client-side script execution (integrity), with no server-side confidentiality or availability effect. No authentication is required to submit the image (PR:N), but the user-interaction and SVG-rendering prerequisites constrain real-world reach.

This is a medium-severity, client-side issue - a real but non-critical priority. The supplied CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:P, base 5.3) indicates a remotely reachable, low-complexity flaw requiring no privileges, but it is gated by passive user interaction (UI:P) and the impact is confined to low integrity (VI:L) with no confidentiality (VC:N) or availability (VA:N) effect - consistent with a cross-site scripting (CWE-79) bug where a sanitization routine fails to strip active content from an SVG. Publicly available exploit code exists (referenced via the imgproxy GitHub issue #1708), which raises the practical likelihood of exploitation, but the flaw is NOT listed in CISA KEV and no EPSS probability was provided in the input, so there is no evidence of widespread or automated active exploitation at this time - POC availability and active exploitation are distinct signals and should not be conflated. An aggravating operational factor is the absence of a vendor fix: the report states the project was notified via an issue but has not responded, so no patched version is confirmed. Net assessment: treat as a moderate-priority hardening item rather than an emergency - the limited impact and user-interaction requirement temper the fact that working exploit code is public and no patch exists. All data derives from a single source (VulDB), which lowers confidence.

An attacker crafts an SVG containing a script/event-handler payload and causes imgproxy to process it through the SVG handler, where the sanitizeElement routine fails to remove the active content; the resulting SVG is served to and rendered by a victim's browser, executing attacker script in the victim's session context. A public exploit is already available (imgproxy GitHub issue #1708), lowering the barrier to reproduction. Exploitation requires the victim to view/render the malicious image (CVSS UI:P), so it is not fully automatic.