Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach MMS model registration endpoint
-
DeliverySubmit crafted model archive
-
ExploitTrigger path traversal during extraction
-
ExecutionWrite file outside model directory
-
ImpactExecute arbitrary code on host
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V5 - File Handling Relative path traversal prevention
|
| Exploitation Conditions | Requires the attacker to supply a crafted model archive to a Multi Model Server instance that extracts it (model registration/load path), exploiting a CWE-23 relative path traversal during extraction to write files outside the model directory and achieve code execution. The CVSS vector states no authentication (PR:N) and no user interaction (UI:N), consistent with MMS's model-registration API accepting archives. The key LIMITING factor is network reachability of that registration/management endpoint: MMS's management API has historically bound to localhost by default, so remote unauthenticated exploitation realistically depends on the endpoint being exposed on a reachable interface or the attacker being able to influence the model source. There is NO patched version (product is end-of-life), so the only complete mitigation per the vendor is to discontinue use and migrate to a supported model-serving solution. |
| Impact | Severity signals are strong and internally consistent: AWS (AMZN) assigned CVSS 4.0 9.3 (Critical) with AV:N/AC:L/AT:N/PR:N/UI:N and high confidentiality, integrity and availability impact, matching a CWE-23 relative path traversal that yields arbitrary code execution. The vuln.today decision label is act_now, reinforcing high urgency. However, several exploitation signals are MISSING rather than negative: no EPSS probability was provided, the input shows no CISA KEV listing (so this is not confirmed as actively exploited), and no PoC or exploit reference is supplied - so there is no public exploit identified at time of analysis. The most important caveat is a likely mismatch between the AV:N/PR:N scoring and real-world reachability: Multi Model Server's management API (used to register model archives) has historically bound to localhost by default, so network-unauthenticated exploitation realistically requires the management/registration endpoint to be exposed to attackers - verify exposure before treating every instance as internet-exploitable. Net assessment: a genuinely high-severity, unauthenticated-RCE-class flaw that is a real priority for anyone still running MMS, but exploitation prevalence cannot be confirmed from the provided data, and the product is end-of-life with NO fixed version available, so risk persists until the product is removed. |
| Exploit ScenarioDerived | An attacker who can reach a Multi Model Server registration/management endpoint supplies a crafted model archive (.mar) whose entries contain relative path traversal sequences; when MMS extracts the archive it writes files outside the intended model directory, overwriting a script or executable that the server later runs, resulting in arbitrary code execution on the host. No public PoC or exploit code was provided in the input, and the attack is rated AC:L (low complexity) with no authentication or user interaction required per the CVSS vector. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv2 – Oct 23, 2026
Relative path traversal during model archive extraction in AWS Multi Model Server 1.1.11 and earlier allows unauthenticated remote actors to write files outside the intended model directory and achieve arbitrary code execution. AWS rates the flaw CVSS 4.0 9.3 (Critical) and has declared the product end of life, so no fixed version will ship; the vendor's guidance is to discontinue use and migrate to a supported model-serving solution. No public exploit code or confirmed active exploitation was identified at time of analysis, and realistic remote exploitation depends on the model registration/load endpoint being reachable rather than bound to localhost.
Earlier versions of this analysis (1)
Path traversal in AWS Multi Model Server 1.1.11 and earlier lets remote, unauthenticated attackers write outside the intended model-extraction directory when a crafted model archive is loaded, enabling arbitrary code execution. The practical prerequisite is that an attacker can get such an archive into the extraction path, typically through a reachable model-registration or management interface that accepts untrusted archives or archive URLs. No public exploit has been identified at the time of analysis, and the priority label is act_now; the product is end-of-life with no fixed version, so AWS advises migration to a supported model-serving solution.
RemediationAI
No vendor-released patch is identified: AWS states Multi Model Server is end of life and directs users to discontinue it and migrate to a supported model-serving solution, which per the vendor is the only complete fix. If migration cannot happen immediately, reduce exposure by keeping the model registration/load endpoint bound to localhost or behind strict network ACLs so untrusted parties cannot submit archives - the trade-off is that this breaks remote model-registration workflows and any pipeline that pushes models over the network - and only load archives from sources you control and have verified. Track the AWS security bulletin for updates: https://aws.amazon.com/security/security-bulletins/2026-148-aws/
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
Multi Model Server (MMS) is AWS's model-serving stack that ingests model archives through its model registration/load path and unpacks them locally before serving them. The root cause is CWE-23 (relative path traversal): extraction of a crafted archive resolves entries outside the intended extraction root, letting an attacker create or overwrite files anywhere the MMS process can write, which is enough to plant code the server subsequently loads and executes. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N with high C/I/A) and the independently assessed CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) agree on unauthenticated network exploitation with no user interaction. The practical limiting factor is reachability of the model registration/management endpoint, which has historically been bound to localhost by default, so remote exploitation realistically requires that endpoint to be exposed on a reachable interface or the attacker to otherwise influence the model source.
Know when Multi Model Server gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
Crafted archive processed by the network-facing registration path with no auth or interaction gives AV:N/AC:L/PR:N/UI:N; code execution on the host yields C:H/I:H/A:H with unchanged scope.
Primary rating from Vendor (AMZN).
CVSS VectorVendor: AMZN
Lifecycle Timeline
3DescriptionCVE.org
Relative path traversal during model archive extraction in Amazon Web Services Multi Model Server 1.1.11 and earlier might allow remote actors to execute arbitrary code via a crafted model archive.
Multi Model Server is end of life and no fixed version is available. Users should discontinue use of Multi Model Server and migrate to a supported model serving solution.
Executive ViewAI
AWS Multi Model Server (MMS) 1.1.11 and earlier contains a critical path traversal flaw that can let unauthenticated remote attackers write files outside the model directory and achieve arbitrary code execution on hosts serving models. Because AWS has declared the product end of life, no patch will ever ship, so any team still running MMS carries a permanent, unresolvable exposure that leadership must treat as a migration and lifecycle risk rather than a routine patch. The stakes rise sharply for any internet-reachable model registration or load endpoint.
No vendor-released patch identified at time of analysis. Interim mitigations: discontinue use and migrate to a supported model-serving solution per vendor guidance; restrict the model registration/load endpoint to localhost or trusted internal networks; apply network segmentation and firewall rules to limit exposure; monitor hosts for unauthorized file writes and unexpected process execution.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Path traversal exploits occur when applications use user-controlled input to construct file system paths without proper validation.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
EUVD-2026-109111
GHSA-3v4f-xjg3-3j32
NVD
CVE.org
EPSS
Exploit-DB