GNU gzip CVE-2026-41992

Exploitation requires the attacker to control two input files processed sequentially within a single gzip -d invocation: the first file must be a specially crafted LZW-format (Unix compress) archive to poison the shared global array, and the second must be a specially crafted LZH-format archive to trigger the out-of-bounds read using the poisoned state - the ordering is mandatory. The CVSS 4.0 vector (AV:L/PR:N) confirms that no elevated privileges are required, but the attacker must be able to run the gzip binary locally or supply files to a backend that does so in batch mode. Exploitation is further limited by: (1) the local attack vector - remote exploitation requires a service intermediary that passes attacker-supplied files to gzip; (2) the specific two-file cross-format sequencing constraint (LZW then LZH in the same invocation); and (3) the absence of reinitialization in the affected build. Single-file gzip decompression invocations are not exploitable by this vulnerability.

The CVSS 4.0 base score of 6.9 reflects a meaningful but constrained vulnerability: local attack vector (AV:L) limits opportunistic exploitation, while low complexity (AC:L), no privileges required (PR:N), and high confidentiality impact (VC:H) indicate that exploitation, when access is available, is straightforward and damaging. The local vector is the primary limiting factor - exploitation cannot occur purely remotely without a backend service that passes attacker-supplied files to gzip in batch. No EPSS probability score, KEV listing, or public proof-of-concept has been provided in the available intelligence, so no public exploit identified at time of analysis. The CVSS 4.0 vector reports VI:N and VA:N, confirming this is a confidentiality-only issue - memory disclosure rather than code execution or denial of service, which reduces severity relative to a typical buffer overflow. Real-world risk is elevated in CI/CD pipelines, file decompression services, or archive processing backends that invoke gzip with untrusted multi-format inputs, where the AV:L constraint is effectively bypassed by the service architecture. Despite gzip's ubiquitous deployment footprint, the two-file cross-format precondition substantially narrows the exploitable attack surface.

An attacker who controls files submitted to a backend service that invokes gzip for archive decompression - such as a CI/CD artifact processor or a file upload decompression endpoint - submits a crafted LZW archive followed by a crafted LZH archive as part of the same batch job. The gzip process decompresses the LZW file first, leaving attacker-controlled stale values in the shared global array, then decompresses the LZH file using those poisoned values, triggering an out-of-bounds read that may expose adjacent heap or stack memory contents to the attacker. No public proof-of-concept has been confirmed at time of analysis, though the CERT.pl disclosure describes the mechanics with sufficient detail to enable reproduction.