Loading vulnerability…
CVE-2026-23001 is a use-after-free vulnerability in the Linux kernel macvlan subsystem. Concurrent source-entry deletion and forwarding can leave macvlan_forward_source() able to observe a macvlan_source_entry whose associated vlan pointer is queued for reclamation. The corrective change applies RCU protection to the vlan pointer and clears that pointer in macvlan_hash_del_source() before the RCU grace period begins, allowing forwarding to skip entries pending destruction.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This repository is a small Python proof-of-concept for insecure deserialization leading to code execution in PyTorch/Hugging Face model-loading workflows. It contains two code files: `malicious_model_card.py`, which generates a fake model package, and `exploit_hf_pickle.py`, which simulates a victim loading the malicious artifact. The core exploit primitive is a Python class with a crafted `__reduce__` method that returns `(os.system, ('id > /tmp/hf_pwned',))`. When the serialized object is later deserialized via `torch.load('pytorch_model.bin')`, the pickle machinery invokes `os.system`, causing command execution. The repository does not include networking or callback infrastructure; it is a local file-based deserialization exploit demonstrating how a malicious model file can act as a supply-chain attack artifact. Structure-wise, `README.md` explains the vulnerability and usage, `malicious_model_card.py` writes `config.json` and a malicious `pytorch_model.bin`, and `exploit_hf_pickle.py` triggers the vulnerable load path. The exploit capability is arbitrary command execution at model load time, with the included payload simply creating `/tmp/hf_pwned` as evidence.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
34 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A Linux kernel possible use-after-free vulnerability in macvlan_forward_source().
A Linux kernel use-after-free vulnerability in macvlan_forward_source() involving RCU protection for struct macvlan_source_entry->vlan.
A Linux kernel macvlan use-after-free vulnerability in macvlan_forward_source fixed by this advisory.
A Linux kernel use-after-free vulnerability in macvlan_forward_source() within the macvlan subsystem.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.