Threat briefs

An unauthenticated remote code execution vulnerability, tracked as CVE-2026-62379, affects OpenAM up to and including version 16.1.1, allowing attackers to achieve full server compromise by sending a crafted XML element to the `/authservice` endpoint that names and instantiates an arbitrary Java class without validation on default configurations.

A critical remote code execution (RCE) vulnerability, CVE-2026-62263, exists in OpenAM's WebAuthn authenticator deserialization, allowing an unauthenticated attacker to bypass an `ObjectInputFilter` and execute arbitrary code by crafting a malicious serialized stream before authentication.

An incomplete fix for a path traversal vulnerability in Open WebUI's terminal proxy allows authenticated attackers to bypass security checks by sending a 9x percent-encoded path, leading to requests being forwarded with terminal credentials and user identification headers to unintended arbitrary paths outside the intended proxy scope.

An authenticated low-privilege user can exploit CVE-2026-59216 in Open WebUI versions prior to 0.10.0 to execute arbitrary Python code or tools within another user's authenticated session by supplying an unvalidated `session_id`, which, if targeting an administrator, leads to remote code execution on the server as the root process.

An authorization bypass vulnerability (CVE-2026-55502) in Cloudreve 4.16.1 allows an attacker with an `Admin.Read` OAuth token to modify the OneDrive storage policy credentials via a POST request to `/api/v4/admin/policy/oauth/signin`, despite lacking `Admin.Write` scope, which can break the storage backend and redirect future OAuth setups.

A high-severity command injection vulnerability, CVE-2026-55404, in yt-dlp versions prior to 2026.7.4 allows remote attackers to achieve arbitrary code execution by crafting malicious metadata that is improperly sanitized when generating Windows `.url` or Linux `.desktop` shortcut files, leading to remote executable execution or shell command injection upon user interaction.

A heap out-of-bounds write vulnerability exists in the vf_quirc filter of FFmpeg versions 7.0 through 8.1.2, allowing an attacker to corrupt heap memory and potentially achieve arbitrary code execution by supplying a crafted PGS/SUP subtitle file with mismatched frame dimensions.

A heap out-of-bounds write vulnerability, CVE-2026-66040, exists in the native PNG and APNG encoders of FFmpeg through version 8.1.2, allowing remote attackers to corrupt heap memory and achieve potential arbitrary code execution by supplying a crafted PNG image with a malicious eXIf chunk.

A signed integer overflow vulnerability, CVE-2026-66039, exists in the MACE6 audio decoder of FFmpeg versions through 8.1.2, allowing attackers to corrupt heap memory and achieve code execution by supplying a crafted CAF file with malicious bytes_per_packet and frames_per_packet values.

A heap out-of-bounds write vulnerability exists in FFmpeg through version 8.1.2, specifically within the vf_hqdn3d filter, allowing attackers to corrupt heap memory by providing a crafted video input where frame resolution increases between frames while filtergraph reinitialization is disabled, leading to undersized buffers and a write beyond allocation boundaries.

A flaw exists in the cluster-proxy service-proxy component of Red Hat Advanced Cluster Management for Kubernetes (RHACM) and multicluster-engine (MCE), allowing an authenticated hub principal to inject an Impersonate-Group header into proxied requests, bypassing validation, and leveraging the spoke ServiceAccount's unrestricted impersonation permissions to escalate privileges to cluster-admin on all managed clusters.

CVE-2026-49952 is an authentication bypass vulnerability in Discuz! X5.0 versions 20260320 through 20260501, allowing unauthenticated remote attackers to gain unauthorized access to database backup and restore functionality by exploiting a shared cryptographic key, leading to potential data exfiltration and user impersonation.

A double-free vulnerability, CVE-2026-66032, in libssh2 versions through 1.11.1 allows a malicious SSH server to corrupt the heap of an authenticated client opening an SFTP session, potentially leading to arbitrary code execution.

sysPass versions up to 3.2.11 are affected by a missing object-level authorization vulnerability in the JSON-RPC API. Attackers holding an API token can exploit this flaw by invoking AccountController methods (e.g., viewAction, editAction, deleteAction, editPassAction) without proper AccountFilterUser checks, allowing them to enumerate account metadata, overwrite passwords, and delete user accounts across the entire vault, bypassing per-account access control defined by their token permissions.

An insecure direct object reference vulnerability (CVE-2026-65708) in sysPass versions up to 3.2.11 allows authenticated attackers to bypass access controls, accessing, enumerating, and manipulating account file attachments by manipulating numeric file IDs in `AccountFileController` actions without proper authorization checks, leading to unauthorized data access.

An authenticated user can overwrite messages in any channel, including private and DM channels, by exploiting the CVE-2026-59714 authorization bypass vulnerability in Open WebUI's chat completion API, leading to message integrity destruction and impersonation.

Open WebUI versions from 0.9.0 to before 0.10.0, when configured with Redis, fail to correctly enforce JWT revocation for realtime authentication endpoints such as Socket.IO and terminal websockets, allowing attackers to maintain access to real-time features with stolen, revoked JWTs.

An unauthenticated remote attacker can cause a Denial of Service (DoS) in Netty servers utilizing XmlFrameDecoder by sending a specially crafted XML payload containing repeated '</' characters, leading to CPU exhaustion of the server's EventLoop thread and unresponsiveness.

A high-severity sandbox escape vulnerability, CVE-2026-55607, exists in Claude Code's worktree handling, allowing attackers to achieve unsandboxed code execution by manipulating symlinks and exploiting Git fsmonitor during worktree operations to overwrite user home directory files like .zshenv, requiring a user to clone a malicious repository and run Claude Code against it.

A stored web worker XSS vulnerability, CVE-2026-59214, in Open WebUI versions prior to 0.10.0 allows a low-privileged user to inject malicious Python code into chat messages that, when executed by an administrator or privileged user via a 'Run' click, triggers authenticated same-origin requests to create server-side functions with arbitrary commands, leading to remote code execution on the Open WebUI server.

An authentication bypass vulnerability (CWE-287) exists in the `openapi3filter.ValidationHandler` component of the `getkin/kin-openapi` library (versions <= v0.143.0), where the `ValidationHandler.Load()` method silently defaults to a `NoopAuthenticationFunc` when an explicit function is not provided, allowing unauthenticated remote attackers to bypass OpenAPI security requirements and access protected endpoints in Go services.

The @fastify/static package is vulnerable to a route guard bypass via path traversal using non-leading '..' or '%2E%2E' path segments, allowing attackers to circumvent route-based middleware and access protected files that are served by the static plugin.

A high-severity Cross-Site Request Forgery (CSRF) bypass vulnerability in React Router's unstable React Server Components (RSC) APIs allows for action execution before a 400 response, impacting applications utilizing these specific APIs.

A denial of service vulnerability exists in the js-yaml library (versions 5.0.0 through 5.2.1) due to an exponential parsing time bug in flow collections, allowing attackers to craft a small YAML document which, when processed by `load()` or `loadAll()` functions, consumes significant CPU resources and blocks the Node.js event loop.

A critical vulnerability in GitPython versions up to 3.1.53 allows attackers to achieve arbitrary command execution by influencing `git clone` options to include a malicious `--template` directory, which causes Git hooks to be copied and executed during cloning, even in default configurations.

A critical vulnerability in GitPython versions up to 3.1.52 allows attackers to inject arbitrary Git configuration directives into a victim's `.git/config` file by crafting a malicious submodule name, leading to remote code execution (RCE) during subsequent Git operations that trigger `core.sshCommand`.

A path traversal vulnerability (GHSA-r28c-9q8g-f849) in PostCSS versions up to 8.5.17 allows an unauthenticated attacker to disclose arbitrary `.map` file contents from a vulnerable server by crafting a `sourceMappingURL` comment in user-submitted CSS, leading to sensitive information exposure if the processed output is returned.

Velocity.js versions up to 2.1.6 are vulnerable to Remote Code Execution (RCE) through an incomplete fix for a previous prototype pollution vulnerability, enabling attackers to craft malicious Velocity templates to leverage unfiltered property-read expressions and execute arbitrary JavaScript code on the server, leading to full server compromise.

A critical server-side template injection vulnerability exists in the @prompty/core Nunjucks renderer, affecting versions <= 0.1.4 and >= 2.0.0-alpha.1 up to <= 2.0.0-beta.4. This flaw allows an attacker to execute arbitrary JavaScript code within the host Node.js process by crafting malicious `.prompty` template bodies. The renderer's unrestricted JavaScript member access permits traversal of constructor and prototype properties, leading to remote code execution when rendering untrusted, community-supplied, cloned, or LLM-generated `.prompty` files.

A broken access control vulnerability in Suna's message queue API allows authenticated attackers to gain unauthorized access to and manipulate queue resources belonging to other users by exploiting missing ownership and account isolation checks. This exploit enables attackers to read all users' pending prompt queues, read or delete individual sessions, and inject arbitrary prompts into another user's session, which causes the background drainer to forward malicious messages to the victim's AI agent using their credentials and permissions, leading to potential data manipulation, unauthorized actions, or further compromise.