Loading vulnerability…
CVE-2024-50050 is a remote code execution vulnerability in the default Meta Reference Python inference implementation of Llama Stack before revision 7a8aa775e5a267cf8660d83140011a0b7f91e005. The implementation accepted objects received over a ZeroMQ socket using pyzmq's recv_pyobj method, which deserialized network-supplied data through Python pickle. Crafted pickle objects can execute code during deserialization, allowing an attacker that can send data to the exposed socket to execute arbitrary code on the Llama Stack inference host. Optional third-party inference integrations are not affected by this default-implementation flaw.
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.
No valid public exploits. Mallory filtered out 1 candidate as fakes, detection scripts, or README-only repos.
All candidate exploits were filtered out by Mallory's validation.
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.
11 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Critical unauthenticated remote code execution vulnerability in Meta Llama Stack's default Meta Reference Python inference implementation. An attacker able to reach the exposed ZeroMQ socket can submit a crafted Python pickle object; recv_pyobj automatically deserializes it with pickle.loads, allowing arbitrary code execution on the inference-server host.
Security flaw in Meta's Llama framework that could allow arbitrary code execution on the llama-stack inference server.
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.