CVE Database

An unchecked return value within the AMD Platform Management Framework (PMF) could allow an attacker to read or modify an arbitrary address potentially resulting in loss of confidentiality, integrity, or availability.

Improper isolation of GPU HW register space could allow a privileged attacker in malicious Guest Virtual Machine (VM) to perform unauthorized access to specific victim range of GPU MMIO register space, potentially causing the host OS to reboot and creating a Denial of Service (DOS) condition.

Improper Input Validation in the AMD RAID driver could allow an attacker to point to an arbitrary memory location potentially resulting in privilege escalation and arbitrary code execution.

Improper restriction of operations within the bounds of a memory buffer in the AMD secure processer (ASP) could allow an attacker to read or write to protected memory potentially resulting in arbitrary code execution.

Improperly preserved integrity of hardware configuration state during a power save/restore operation in the AMD Secure Processor (ASP) could allow an attacker with the ability to write outside the trusted memory range (TMR) to change the execution flow of the Video Core Next (VCN) firmware potentially impacting confidentiality, integrity, or availability.

Improper validation in Power Management Firmware (PMFW) may allow an attacker with privileges to pass malformed workload arguments when exporting table data from SMU to DRAM potentially resulting in a loss of confidentiality and/or availability.

A TOCTOU (Time-Of-Check to Time-Of-Use) in the graphics interface may allow an attacker to load registers repeatedly creating a race condition potentially leading to a loss of integrity.

A compromised Trusted OS (TOS) driver could issue a malformed call that could potentially allow memory access outside the intended range resulting in loss of integrity.

WWW::Mechanize::Cached versions before 2.00 for Perl deserialize cached HTTP responses from a world-writable on-disk cache, enabling local response forgery and code execution. With no explicit cache backend, WWW::Mechanize::Cached constructs a default Cache::FileCache under /tmp/FileCache without overriding the backend's documented directory_umask of 000, so the cache root and its subdirectories are created mode 0777 with no sticky bit. Cache entries are named by sha1_hex of the request and read back through Storable::thaw on the next cache hit. A local attacker with write access to the cache tree can replace a victim's cache entry for a known URL with an arbitrary frozen HTTP::Response blob, causing the victim's next get() of that URL to return attacker controlled response bytes. Because the bytes are passed to Storable::thaw, a victim process that has loaded any class with a side-effectful STORABLE_thaw, DESTROY, or overload hook can be escalated to arbitrary code execution.

A System Management Mode (SMM) handler could perform a callout to code located in non-SMM/untrusted memory. A highly privileged attacker could, with active user interaction and under high complexity and present preconditions, trigger execution of attacker-controlled code in SMM, potentially compromising the system’s confidentiality, integrity, and availability.

Incorrect default permissions in the installation directory for the AMD chipset driver could allow an attacker to achieve privilege escalation resulting in arbitrary code execution.

An improper input validation vulnerability within the AMD Platform Management Framework (PMF) Driver can allow a local attacker to write Out-of-Bounds, potentially resulting in privilege escalation.

Improper input validation in the AMD Secure Processor (ASP) PCI driver could allow a local attacker to trigger a Use-After-Free (UAF) condition, potentially resulting in a loss of platform integrity or crash.

An improper input validation vulnerability within the AMD Platform Management Framework (PMF) driver can allow a local attacker to read Out-of-Bounds potentially resulting in information disclosure or a crash

An improper input validation vulnerability within the AMD Platform Management Framework (PMF) driver can allow a local attacker to read or write Out-of-Bounds, potentially resulting in privilege escalation

Incorrect default permissions in the installation directory for the AMD general-purpose input/output controller (GPIO) could allow an attacker to achieve privilege escalation resulting in arbitrary code execution.

Improper Input validation in the AMD Secure Processor (ASP) PCI driver may allow a local attacker to create a buffer overflow condition, potentially resulting in a crash or denial of service

Improper input validation in the AMD OverDrive (AOD) System Management Mode (SMM) module could allow a privileged attacker to perform an out-of-bounds read, potentially resulting in loss of confidentiality.

Stack exhaustion vulnerability in the MongoDB PHP driver can cause application crashes when processing deeply nested BSON documents in unusual circumstances when the source of these BSON documents is not MongoDB Server.

Hedera Guardian through 3.5.1 contains an authentication bypass vulnerability in the GET /api/v1/demo/registered-users endpoint that allows unauthenticated attackers to retrieve sensitive user information. Attackers can access the endpoint without providing authentication credentials to obtain usernames, Hedera DIDs, parent registry DIDs, system roles, and policy role assignments for all registered users in the system.

ZITADEL is an open source identity management platform. From 2.71.11 to before 3.4.10 and 4.15.0, a vulnerability was discovered in Zitadel's LDAP identity provider implementation, which fails to properly escape user-provided usernames before incorporating them into LDAP search filters. This allows unauthenticated attackers to perform LDAP Filter Injection during the login process. While this vulnerability does not allow for a full authentication bypass, an attacker can use LDAP metacharacters (such as *, (, )) to perform blind LDAP injection. By observing the different failure (or success) responses, an attacker can systematically enumerate valid usernames and extract sensitive attribute data from the connected LDAP directory. This vulnerability is fixed in 3.4.10 and 4.15.0.

The MCP Registry provides MCP clients with a list of MCP servers, like an app store for MCP servers. Prior to 1.7.6, the client-side and server-side GitHub OIDC flow is bound only to a global audience string, not to the specific registry instance being targeted. On the client side, the publisher always appends audience=mcp-registry when requesting the GitHub Actions ID token, regardless of the selected --registry URL. On the server side, the exchange endpoint validates only that same fixed audience and then derives publish permissions directly from repository_owner. As a result, a token legitimately obtained while interacting with one registry deployment remains acceptable to any other deployment that shares the same code and audience string. This vulnerability is fixed in 1.7.6.

The MCP Registry provides MCP clients with a list of MCP servers, like an app store for MCP servers. From 1.1.0 to 1.7.4, the TrailingSlashMiddleware in internal/api/server.go is vulnerable to an open redirect attack. An attacker can craft a URL with a protocol-relative path (e.g., //evil.com/) that, after trailing slash removal, results in a Location header of //evil.com — which browsers interpret as an absolute URL to an external domain. This vulnerability is fixed in 1.7.5.

The MCP Registry provides MCP clients with a list of MCP servers, like an app store for MCP servers. Prior to 1.7.9, OCI ownership validation skips label-match check when upstream OCI registry returns HTTP 429, letting any authenticated publisher bind their io.github.<user>/* namespace to OCI images they do not control. internal/validators/registries/oci.go:104-119 fails open on http.StatusTooManyRequests: when the registry's anonymous fetch to the upstream OCI registry is rate-limited, ValidateOCI returns nil and the publish is accepted without ever running the io.modelcontextprotocol.server.name label-match check at lines 122-141. That label check is the only cross-system ownership proof the registry applies to OCI packages — every other registry type (NPM, PyPI, NuGet, MCPB) treats a non-200 upstream response as a hard error. This vulnerability is fixed in 1.7.9.

python-utcp is the python implementation of UTCP. Prior to 1.1.3, _prepare_environment() in cli_communication_protocol.py passes a full copy of os.environ to every CLI subprocess. When combined with CVE-2026-45369, an attacker can exfiltrate all process-level secrets in a single tool call. This vulnerability is fixed in 1.1.3.

python-utcp is the python implementation of UTCP. Prior to 1.1.3, the _substitute_utcp_args method in cli_communication_protocol.py inserts user-controlled tool_args values directly into shell command strings without any sanitization or escaping. These commands are then executed via /bin/bash -c (Unix) or powershell.exe -Command (Windows), allowing an attacker to inject arbitrary shell commands. This vulnerability is fixed in 1.1.3.

Elixir WebRTC is an Elixir implementation of the W3C WebRTC API. Prior to 0.15.1 and 0.16.1, missing DTLS peer certificate fingerprint validation in the DTLS client (active) role removes one side of WebRTC's mutual authentication. The bug is not independently exploitable for media interception in standard deployments, but enables a full man-in-the-middle attack when chained with insecure signalling or a peer with similar validation gaps. This vulnerability is fixed in 0.15.1 and 0.16.1.

Tuist is a virtual platform team for Swift app devs. Prior to 1.180.10, the forgot password flow allows an unauthenticated attacker to repeatedly trigger password reset emails for a known account without server-side throttling. In self-hosted deployments, this can be abused to send large volumes of unwanted email and consume downstream email delivery resources. This vulnerability is fixed in 1.180.10.

Tuist is a virtual platform team for Swift app devs. In 1.180.8 and earlier, the DELETE /api/projects/{account_handle}/{project_handle}/previews/{preview_id} endpoint loads the preview by its UUID without verifying that the preview belongs to the project resolved from the URL path. The route's project-level authorization plug (AuthorizationPlug, :preview) authorizes the caller against the project encoded in account_handle/project_handle — which the attacker controls — and then the action deletes whichever preview's UUID is supplied. The check therefore guards the wrong project.

libyang is a YANG data modeling language library. Prior to SO 5.2.15, lyb_read_string() in src/parser_lyb.c contains an integer overflow that results in a heap buffer overflow when parsing a maliciously crafted LYB binary blob. An attacker who can supply LYB data to any libyang consumer (NETCONF server, sysrepo, etc.) can trigger a crash or potential heap corruption. This vulnerability is fixed in SO 5.2.15.