Ghidra CVE-2026-115396

Exploitation requires a victim to use Ghidra's theme import feature (readZipTheme/ThemeReader.processIconFile) to import an attacker-supplied theme ZIP containing a highly compressed icon entry in the image/ subdirectory, which is extracted to ~/.config/ghidra/Ghidra*/images without a decompressed-size or compression-ratio limit. Limiting factors: the attack is local (AV:L) and needs user interaction (UI:P) - the user must be socially engineered into importing the crafted archive; there is no remote or network-triggered path. Impact is confined to availability (filesystem/disk and I/O exhaustion causing later writes and other applications to fail); there is no code execution, privilege gain, or data disclosure. Affected versions are Ghidra >= 10.3; the only complete fix stated by the vendor is updating to 12.2, which adds SecureZipExtractor enforcing a 100:1 max compression ratio, 100 MB per-file and 1 GB total extraction limits.

This is a medium-severity availability-only flaw (CVSS 4.0 base 6.8), not a high-priority emergency. The input's own signals are internally consistent: the CVSS 4.0 vector (AV:L/AC:L/PR:N/UI:P, VA:H with VC:N/VI:N) describes a local, low-complexity, decompression-bomb denial of service that requires the victim to perform an action (importing a theme) and produces no confidentiality or integrity impact. The vuln.today decision label is 'this_month' rather than 'act_now' or 'emergency', which matches the limited, local, interaction-dependent nature of the bug. No EPSS score, no CISA KEV listing, and no PoC/exploit reference were supplied, so there is no evidence of active exploitation or of weaponized exploit code. The only material conflict is metadata: the single Tag 'Information Disclosure' directly contradicts the CVSS vector (VC:N, VA:H) and the NSA GHSA advisory, both of which characterize this as uncontrolled resource consumption / denial of service - the tag should be read as mislabeled, not as evidence of data exposure. Net: a legitimate but bounded issue best remediated on the normal patch cycle rather than treated as an exploited, widespread threat.

An attacker crafts a small theme ZIP in which an icon entry under the image/ subdirectory decompresses to an enormous size, then persuades a Ghidra user (e.g. via a shared 'custom theme') to import it. Because the vector is local and requires user interaction (AV:L/UI:P), there is no remote or unattended path; the victim must actively import the file. No public exploit code is identified at time of analysis, though the NSA GHSA advisory documents the mechanism and shows a fully consumed disk (/dev/sda2 at 100%).