Sdl Image CVE-2026-35444

Victim must open a crafted .xcf image file in an application using SDL_image library (any version containing do_layer_surface() function in src/IMG_xcf.c). No authentication required. Affects indexed color XCF files with undersized colormaps and out-of-range pixel indices.

The CVSS v3.1 score of 7.1 (High) reflects network-based attack vector with low complexity and no authentication requirements, though user interaction is required (AV:N/AC:L/PR:N/UI:R). The impact profile shows high confidentiality impact but no integrity impact and low availability impact (C:H/I:N/A:L), consistent with an information disclosure vulnerability. Real-world risk is moderate to significant because SDL_image is embedded in numerous applications including games, image viewers, and content management systems where users routinely open untrusted image files. The heap memory disclosure of up to 762 bytes could leak cryptographic keys, session tokens, or other sensitive data depending on allocator state. However, successful exploitation requires social engineering to convince users to open crafted XCF files, and the leaked data appears in rendered output rather than being directly exfiltrated, reducing immediate severity. No public exploit identified at time of analysis and no CISA KEV listing indicates this is not yet actively exploited in the wild. The fixed commit (996bf12888925932daace576e09c3053410896f8) is available, suggesting patches should be readily deployable. Attack automation is possible but requires targeted file distribution. Technical impact is limited to information disclosure rather than code execution, positioning this as a priority for applications processing untrusted XCF files but not requiring emergency response for most deployments.

An attacker crafts a malicious XCF file with a deliberately small colormap (e.g., 16 entries) but embeds tile data containing pixel index values of 250 or higher. When a victim opens this file in any application using vulnerable SDL_image (such as a game level editor, image viewer, or content creation tool), the do_layer_surface() function reads 762 bytes beyond the colormap heap allocation. These out-of-bounds bytes, which may contain cryptographic material, authentication tokens, or other sensitive data from adjacent heap objects, are copied into the output surface pixel buffer. The attacker retrieves the rendered image through normal application workflows (saved files, screenshots, or network transmission in collaborative tools) and extracts the leaked heap data from the pixel values. No public exploit identified at time of analysis, but the attack requires only file format manipulation skills and standard image editing tools to construct the malicious payload.