{"id":"CVE-2026-41486","aliases":["GHSA-mw35-8rx3-xf9r","PYSEC-2026-2272","PYSEC-2026-2296"],"title":"Ray: Remote Code Execution via Parquet Arrow Extension Type Deserialization","summary":"Ray: Remote Code Execution via Parquet Arrow Extension Type Deserialization","severity":"high","vendor":"ray","product":"ray","ecosystem":"pip","affected":["ray >= 2.49.0, < 2.55.0"],"patched":["ray 2.55.0"],"published":"2026-04-24","updated":"2026-09-10","sourceUpdated":"2026-09-10T03:51:02.380215690Z","source":"OSV","sourceUrl":"https://osv.dev/vulnerability/GHSA-mw35-8rx3-xf9r","references":[{"url":"https://github.com/ray-project/ray/security/advisories/GHSA-mw35-8rx3-xf9r"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-41486"},{"url":"https://github.com/ray-project/ray/pull/54831"},{"url":"https://github.com/ray-project/ray/pull/62056"},{"url":"https://github.com/ray-project/ray/commit/c02bd31ae31996805868baa446a131a8d304525f"},{"url":"https://github.com/ray-project/ray"},{"url":"https://github.com/ray-project/ray/releases/tag/ray-2.55.0"}],"tags":["osv","pip"],"epss":0.00473,"epssPercentile":0.40076,"ingestedAt":"2026-07-13T18:58:00.874Z","slug":"CVE-2026-41486","body":"## Overview\n\nRay Data registers custom Arrow extension types (`ray.data.arrow_tensor`, `ray.data.arrow_tensor_v2`, `ray.data.arrow_variable_shaped_tensor`) globally in PyArrow. When PyArrow reads a Parquet file containing one of these extension types, it calls `__arrow_ext_deserialize__` on the field's metadata bytes. Ray's implementation passes these bytes directly to `cloudpickle.loads()`, achieving arbitrary code execution during schema parsing, before any row data is read.\n\nIn May 2024, Ray fixed a related vulnerability in `PyExtensionType`-based extension types ([issue #41314](https://github.com/ray-project/ray/issues/41314), [PR #45084](https://github.com/ray-project/ray/pull/45084)). In July 2025, [PR #54831](https://github.com/ray-project/ray/pull/54831) introduced `cloudpickle.loads()` into the replacement extension types' deserialization path, reintroducing the same class of vulnerability.\n\n## Impact\n\n- **Affected versions**: Ray 2.49.0 through 2.54.0 (latest release as of March 2026). The vulnerable `_deserialize_with_fallback` function with `cloudpickle.loads()` was introduced in commit `f6d21db1a4` ([PR #54831](https://github.com/ray-project/ray/pull/54831), July 2025), first released in Ray 2.49.0.\n- **Affected configurations**: Any process that uses Ray Data and reads Parquet files. The extension types are registered globally in PyArrow, so all Parquet reads in the process are affected, including `ray.data.read_parquet()`, `pyarrow.parquet.read_table()`, `pandas.read_parquet()`, etc.\n- **Attacker prerequisites**: The attacker must place a crafted Parquet file where a Ray Data pipeline reads it. No authentication or cluster access is required. The Parquet file must contain a column with a `ray.data.arrow_tensor` (or v2, or variable-shaped) extension type name, which makes this a targeted attack against Ray Data users.\n\n## Affected packages\n\n- `ray >= 2.49.0, < 2.55.0`\n\n## Remediation\n\nUpgrade to a patched release:\n\n- `ray 2.55.0`","depth":"twilight","depthScore":41,"depthScoreParts":{"impact":41.3,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}