CVE-2026-25087 is a high-severity (CVSS 7) Use After Free vulnerability in pyarrow. A fix is available for pyarrow — see the affected versions and patch details below.
Apache Arrow: Potential use-after-free when reading IPC file with pre-buffering
Exploitation Status
No confirmed exploitation observed yet
- CISA’s own triage has not observed active exploitation or public proof-of-concept code for this CVE as of its last assessment.
Exploitation and automatability from CISA’s SSVC triage for CVE-2026-25087.
EPSS Exploitation Probability
EPSS (Exploit Prediction Scoring System) is a daily probability model maintained by FIRST.org. It estimates the likelihood a CVE will be exploited in production environments within the next 30 days, derived from real-world threat intelligence signals.
How urgent is this, really
CVE-2026-25087 plotted by exploitation likelihood (EPSS) against impact (CVSS). The shaded corner — EPSS 50%+ and CVSS 7.0+ — is where this CVE doesn't sit, though severity or exploitability alone can still warrant action.
Where this sits among everything scored
Of 377,636 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Real counts from FIRST.org, not a sample — log-scaled since the landscape is heavily right-skewed.
Real-World Exposure
pyarrowReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.
Description
Use After Free vulnerability in Apache Arrow C++.
This issue affects Apache Arrow C++ from 15.0.0 through 23.0.0. It can be triggered when reading an Arrow IPC file (but not an IPC stream) with pre-buffering enabled, if the IPC file contains data with variadic buffers (such as Binary View and String View data). Depending on the number of variadic buffers in a record batch column and on the temporal sequence of multi-threaded IO, a write to a dangling pointer could occur. The value (a std::shared_ptr<Buffer> object) that is written to the dangling pointer is not under direct control of the attacker.
Pre-buffering is disabled by default but can be enabled using a specific C++ API call (RecordBatchFileReader::PreBufferMetadata). The functionality is not exposed in language bindings (Python, Ruby, C GLib), so these bindings are not vulnerable.
The most likely consequence of this issue would be random crashes or memory corruption when reading specific kinds of IPC files. If the application allows ingesting IPC files from untrusted sources, this could plausibly be exploited for denial of service. Inducing more targeted kinds of misbehavior (such as confidential data extraction from the running process) depends on memory allocation and multi-threaded IO temporal patterns that are unlikely to be easily controlled by an attacker.
Advice for users of Arrow C++:
-
check whether you enable pre-buffering on the IPC file reader (using
RecordBatchFileReader::PreBufferMetadata) -
if so, either disable pre-buffering (which may have adverse performance consequences), or switch to Arrow 23.0.1 which is not vulnerable
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | pyarrow | ≥ 15.0.0&&< 23.0.1 | 23.0.1pip install --upgrade 'pyarrow==23.0.1' |
Detection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for pyarrow, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update pyarrow to 23.0.1 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms CVE-2026-25087 is resolved across your whole dependency graph.
Workarounds
If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.
How O3 protects you
O3 Security's impact-aware SCA analyses which vulnerable code paths your application actually calls, so a match like CVE-2026-25087 can be triaged on real exposure rather than presence alone.
Tailored to CVE-2026-25087. Runtime protection reduces exposure until a permanent patch is applied and verified — it complements patching, it doesn't replace it.
Fixing This On Your OS
If you run this on a Linux distribution, patch through your package manager against the distro's own security advisory below — it tracks the exact backported fix for your release, which can ship on a different timeline (and sometimes a different severity) than the upstream project.
| Product | Fixed in | Advisory |
|---|---|---|
| Red Hat AI Inference Server 3.2 | rhaiis/model-opt-cuda-rhel9:1787772157 | RHSA-2026:61628 |
Frequently Asked Questions
Is CVE-2026-25087 in your dependencies?
O3 Security finds CVE-2026-25087 across PyPI dependencies, including transitive ones, and its impact-aware SCA ranks findings by whether your code actually calls the vulnerable path.