CVE-2026-84452
HighCVSS 8.6Exploitation Probability (EPSS)
Elevated risk75th percentile - higher than 75% of all known CVEs
Summary
Windows ML CLI before version 0.4.0 exposes commands through a localhost HTTP API without authentication and with wildcard allow_origins. A malicious website can send requests to /v1/cli/build or /v1/cli/config with trust_remote_code=true, leading to import of Python code from an attacker-controlled repository and arbitrary code execution as the server user.
Risk Assessment
Visiting a malicious page by a user can result in arbitrary code execution on the server, potentially leading to full system compromise.
Recommendation
Update Windows ML CLI to version 0.4.0 or later, which contains the fix.
Related vulnerabilities
- CVE-2026-123614Critical
imgproxy from 4.0.0 before 4.0.16 contains an integer overflow in the ICO loader that allows remote attackers to trigger a heap buffer overflow via a crafted data_size value. Attackers can submit a crafted ICO image whose data_size wraps the allocation in vips/icoload.c, corrupting heap memory to crash the process or achieve code execution.
- CVE-2026-123551Critical
OpenAM from 14.0.0 before 16.1.4 contains an information disclosure vulnerability that allows unauthenticated attackers to read live session tokens from the notification WebSocket. Attackers can bypass the agent-only check through an unvalidated internal forward on legacy Liberty ID-FF endpoints, collecting iPlanetDirectoryPro tokens to hijack user and administrator sessions.
- CVE-2026-123550Critical
OpenAM before 16.1.4 contains an authentication bypass vulnerability in its SAML v2.0 service provider that accepts unsigned assertions on the ECP (PAOS) and HTTP-Artifact bindings when WantAssertionsSigned is false. Unauthenticated remote attackers can submit forged assertions to obtain sessions as any mapped user, including administrators, and hijack single sign-on to dependent applications.
- CVE-2026-123549Critical
OpenAM before 16.1.4 fails to enforce the Valid Creation Attributes list during self-registration, allowing anonymous attackers to set restricted attributes including group membership. Attackers can register through the legacy self-service REST endpoint or User Self-Service flow to join administrative groups and gain full administrative control of OpenAM.
- CVE-2026-124060Critical
A vulnerability was detected in TOTOLINK CA300-PoE version 6.2c.884 in the UploadCustomModule function of the file /cgi-bin/cstecgi.cgi. Manipulation of the File argument results in a buffer overflow. The attack can be launched remotely, and the exploit is now public and may be used.
- CVE-2026-124059Critical
A security vulnerability has been detected in TOTOLINK CA300-PoE 6.2c.884. The function setWiFiWdsAddConfig in the file /cgi-bin/cstecgi.cgi is affected, where manipulation of the ApCliKeyStr argument leads to a buffer overflow. The attack may be initiated remotely, and the exploit has been disclosed publicly and may be used.
- CVE-2026-124058Critical
A weakness has been identified in TOTOLINK CA300-PoE 6.2c.884. The function setWiFiRepeaterConfig in the file /cgi-bin/cstecgi.cgi is affected, where manipulation of the ApCliKeyStr argument can lead to a buffer overflow. The attack can be launched remotely, and the exploit has been made available to the public and could be used for attacks.
- CVE-2026-124057Critical
A security flaw has been discovered in TOTOLINK CA300-PoE 6.2c.884. This impacts the function setWiFiMultipleConfig of the file /cgi-bin/cstecgi.cgi. Manipulation of the argument Key1Str1 results in a buffer overflow. The attack can be initiated remotely. The exploit has been released to the public and may be used for attacks.
- CVE-2026-123966Critical
A vulnerability was found in Tenda AC19 version 16.03.10.07 in the saveParentControlInfo function of the /goform/saveParentControlInfo file. Manipulating the deviceId argument causes a heap-based buffer overflow. The attack can be carried out remotely, and the exploit has been publicly disclosed.
- CVE-2026-120984Critical
A flaw was found in the Keycloak server's client management component. When a client is created or imported with authorization services enabled, the system fails to properly clear an internal session attribute used during setup. An authenticated attacker with client management permissions can bypass security controls and upload a malicious JavaScript policy that executes on the server.
Original NVD description (English source)
Windows ML CLI is a command line tool for building portable, performant, and high-quality AI models for Windows ML. Prior to 0.4.0, the src/winml/modelkit/serve/cli_api.py component exposes WinML CLI commands through a localhost HTTP API without authentication and configures the allow_origins setting as a wildcard in both src/winml/modelkit/serve/cli_api.py and src/winml/modelkit/serve/app.py. A malicious website loaded by a user can send cross-origin requests to /v1/cli/build or /v1/cli/config and set the trust_remote_code parameter to true, which is converted to the --trust-remote-code command-line flag without validation. This reaches AutoConfig.from_pretrained with trust_remote_code=True in src/winml/modelkit/loader/_autoconfig.py and imports Python code from an attacker-controlled model repository, resulting in arbitrary code execution as the server user. This issue is fixed in version 0.4.0.
Vulnerability data from NVD (NIST) · CISA KEV · EPSS

