Microsoft security update for CVE-2025-12058
🔗 CVE IDs covered (1)
📋 Microsoft's advisory
EchelonGraph does not copy Microsoft's advisory page: its FAQ and remediation text are not shown here. Read the advisory on https://msrc.microsoft.com/update-guide/vulnerability/CVE-2025-12058. This page also shows what EchelonGraph holds for CVE-2025-12058, with the description from its CVE record, which the CVE Numbering Authority that assigned the CVE wrote: for a CVE Microsoft assigned, often in the words of Microsoft's own advisory.
🎯 Affected products6
- azl3 keras 3.3.3-4 on Azure Linux 3.0
- azl3 keras 3.3.3-5 on Azure Linux 3.0
- azl3 keras 3.3.3-6 on Azure Linux 3.0
- azl3 keras 3.3.3-7 on Azure Linux 3.0
- cbl2 keras 2.11.0-3 on CBL Mariner 2.0
- cbl2 keras 2.11.0-3 on CBL-Mariner 2.0
✅ Updates Microsoft lists: 1 update
1 row: one per KB article or update, update type and fixed build. Their details are on Microsoft's advisory.
| KB article | Update | Fixed build |
|---|---|---|
| — | Security Update | — |
🧭 EchelonGraph on CVE-2025-12058
- EchelonGraph score
- 5.9 MEDIUM
- CISA KEV
- Not listed
- Fix status
- A fix is available — apply it.
Description from the CVE record (CVE.org), as published by its numbering authority
The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF). This vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path. * Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer's configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system. * Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server's behalf, resulting in an SSRF condition. The security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access.
Published by Google LLC, the CVE Numbering Authority that assigned CVE-2025-12058, in its CVE record at CVE.org; reused under the CVE Program's Terms of Use (its licence and MITRE's copyright notice are in our Terms of Service).