External risk intelligence

AWS Multi Model Server Archive Extraction RCE Vulnerability.

CVE advisorySeverity: CRITICAL (CVSS 9.3)

CVE-2026-115812

Amazon Web Services Multi Model Server is designed to serve machine learning models via web APIs. As a model serving framework, it is typically deployed as an externally reachable service or API endpoint to facilitate model inference, placing its extraction and processing functions in a position often exposed to network-based input.

Path Traversal

Halo Surface Signal: 4 out of 5 — likely to be public-facing.

External exposure likelihood

Horizon Alert

Summary of the vulnerability and why it matters

A vulnerability in the Multi Model Server, an end-of-life Amazon Web Services product, could allow attackers to execute code remotely by exploiting how model archives are processed. As this product is no longer supported, the primary concern is to confirm if it is in use and to migrate to a modern, supported solution.

  • Unsanitized archive processing allows code execution.
  • Product is end-of-life, migration is essential.
  • Confirm usage and migrate to a supported solution.

Attack Path

How an attacker could exploit the issue

An attacker could target the model archive extraction feature of Amazon Web Services Multi Model Server to gain code execution. This is possible because the server processes model archives, and a specially crafted archive could exploit a path traversal vulnerability to overwrite system files. Since the server is end-of-life, there is no direct fix, and migration to a supported solution is advised.

  • Exposed server accepts crafted archives.
  • Path traversal in extraction triggers vulnerability.
  • Arbitrary code execution is the risk.

Live Threat

Current exploitation, exposure, and threat context

A remote actor could execute arbitrary code by uploading a specially crafted model archive to Amazon Web Services Multi Model Server when supported by the advisory. This could affect the integrity and availability of the model serving instance.

  • Arbitrary code execution.
  • Crafted model archive upload.
  • Compromised model serving instance.

Operational Fix

Recommended remediation, mitigation, and detection steps

The critical vulnerability in Amazon Web Services Multi Model Server impacts organizations using this end-of-life model serving solution. Owners of applications and infrastructure hosting Multi Model Server instances must identify their deployments, confirm external reachability and business criticality, and coordinate with vendor management to plan migration to a supported alternative, as no direct fix is available.

  • Application and infrastructure owners should act.
  • Verify Multi Model Server deployment and reachability.
  • Plan migration to a supported solution.

Supplementary metadata

Validate whether this threat affects your internet-facing exposure.

Halo Threat Intelligence helps prioritize remediation with Halo Surface Signal and H/A/L/O context. Start exposure validation with a free external attack surface trial.

Frequently asked questions

What is Amazon Web Services Multi Model Server?

It is a framework designed to deploy and serve machine learning models through web APIs. Developers use it to make inference capabilities accessible to applications, often running it as an active service that processes incoming data or model files. Because it manages the lifecycle of these models, it must handle incoming archive files, which is the specific component affected by this security issue.

How does CVE-2026-115812 cause a security risk?

This vulnerability is classified as a relative path traversal (CWE-23). It occurs when the server extracts model archives without properly validating file paths. An attacker can craft a malicious archive that directs the software to save files outside of the intended directory. This allows the attacker to overwrite critical system files and eventually achieve arbitrary code execution on the host machine.

Do I need to upload a model to trigger this bug?

Yes, the vulnerability is triggered by the processing of a crafted model archive. It is not triggered by standard API queries or general network traffic that does not involve the archive extraction process. The risk specifically manifests when the server attempts to unpack a file that has been deliberately manipulated to exploit the path traversal flaw during the extraction routine.

Is my server at risk if it is not internet-facing?

Halo Surface Signal notes that Multi Model Server is often deployed as an externally reachable API endpoint, which increases the likelihood of attack. However, even if your instance is internal, the vulnerability remains a significant risk if an attacker gains access to your network. Any entity—internal or external—capable of interacting with the model archive upload function can potentially exploit this weakness.

How should I respond to this vulnerability?

Since Amazon Web Services Multi Model Server has reached its end-of-life, no official security patch will be provided to fix this flaw. Your primary action should be to audit your infrastructure to locate any remaining instances of this software. Once identified, you should prioritize migrating your machine learning workflows to a modern, supported model-serving alternative as soon as possible to ensure ongoing security.

References