libmodulemd EUVD-2026-107036

Exploitation requires a libmodulemd build compiled WITH rpmio support enabled, and a code path that decompresses compressed modulemd YAML through the custom read handler that omits a cumulative decompressed-size check before parsing. The attacker must be able to supply or influence the compressed YAML file that this parser consumes. Limiting factors: the CVSS attack vector is LOCAL (AV:L), so the input must reach the parser on the host rather than being delivered remotely by default; builds without rpmio support are not affected; and impact is limited to availability (service hang/crash) with no confidentiality or integrity compromise. No authentication or user interaction is required once a malicious file reaches the parser (PR:N/UI:N).

This is a moderate-severity availability-only issue, correctly scored CVSS 6.2 (CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) by Red Hat. The CVSS vector is the dominant signal: attack vector is LOCAL (AV:L), meaning the attacker must be able to deliver a malicious compressed modulemd YAML file into a parsing context on the host rather than reach it over the network - this substantially limits real-world exploitability compared to a network-facing flaw. Complexity is low and no privileges or user interaction are required (AC:L/PR:N/UI:N), but impact is confined entirely to availability (C:N/I:N/A:H) via a classic data-amplification/decompression-bomb pattern (CWE-409). No EPSS score, no CISA KEV listing, and no public proof-of-concept were provided in the input, so there is no public exploit identified at time of analysis and no evidence of active exploitation. Signals are consistent rather than conflicting: a low-privilege but local-only DoS with no confidentiality/integrity impact is a genuine hardening priority for systems that ingest untrusted module metadata, but it is not an emergency and ranks below network-reachable or code-execution flaws. Note the AV:L designation should be verified against the actual deployment: if any workflow allows remote or untrusted parties to feed modulemd content to the parser (e.g., a build/repo service), the effective exposure is higher than the local vector implies.

An attacker able to place content where libmodulemd will parse it (for example a repository or build pipeline that ingests module metadata) supplies a highly compressible compressed YAML file. When a build with rpmio support decompresses it through the custom read handler without any cumulative size cap, the small input expands into a huge decompressed stream, exhausting CPU and memory and crashing or hanging the consuming process. No public proof-of-concept was identified in the provided data.