In Splunk AI Toolkit versions below 5.7.3, a low-privileged user that does not hold the 'admin' or 'power' roles could access confidential data that was restricted through `srchFilter` configurations on custom roles.<br><br>The app contains an `authorize.conf` configuration file with a `srchFilter` entry that modifies the built-in ‘user’ role. Because the Splunk platform combines inherited search filters with the `OR` SPL operator, the injected filter overrides more restrictive filters on child roles.

splunk

NVIDIA TRT-LLM for any platform contains a vulnerability in RPC testing, where an attacker could cause an unsafe deserialization. A successful exploit of this vulnerability might lead to code execution, denial of service, data tampering, and information disclosure.

nvidia

NVIDIA TRT-LLM for any platform contains a vulnerability where an attacker could cause an unchecked return value to a null pointer dereference. A successful exploit of this vulnerability might lead to denial of service.

nvidia

NVIDIA TRT-LLM for any platform contains a deserialization vulnerability and unsafe serialized handle. A successful exploit of this vulnerability might lead to code execution, data tampering, and information disclosure.

nvidia

NVIDIA TRT-LLM for any platform contains a vulnerability in MPI server, where an attacker could cause an unsafe deserialization. A successful exploit of this vulnerability might lead to code execution, denial of service, data tampering, and information disclosure.

nvidia

Trilium Notes is a cross-platform, hierarchical note taking application focused on building large personal knowledge bases. In versions 0.102.1 and prior, the Electron configuration is vulnerable to TCC Bypass via Prompt Spoofing, allowing local attackers to trigger misleading macOS permission prompts by running malicious code under the identity of the trusted app. The root cause is that the RunAsNode fuse allows launching the app in a special Node.js mode using -e to execute arbitrary system commands with Trilium Notes's permissions and identity. An attacker can leverage this through a subprocess to request any sensitive permissions, such as access to hardware (camera, microphone) and TCC-protected files, causing the TCC system prompt to appear as if the request came from Trilium rather than the attacker's code, because macOS treats the subprocess as part of the parent application. Exploitation allows access to TCC-protected resources like the screen, camera, microphone, and folders such as ~/Documents and ~/Downloads, undermining macOS's security model and UI integrity through social engineering. This issue has been fixed in version 0.102.2.

Trilium Notes is an open-source, cross-platform hierarchical note taking application for building large personal knowledge bases. Versions 0.102.1 and prior are vulnerable to Local File Inclusion, allowing an authenticated attacker to read sensitive arbitrary files from the server's filesystem. The uploadModifiedFileToAttachment function, which is called when a POST request is received to /api/attachments/{attachmentId}/upload-modified-file, replaces the content of the attachment with the content from another file (whose path is provided in filePath of Request body). After which the content of the attachment can be viewed at /api/attachments/{attachmentId}/download. This exposes sensitive system files such as SSH keys, credentials, configs, and OS files, potentially leading to remote code execution and compromise of co-hosted applications. This issue has been fixed in version 0.102.2.

Discourse is an open-source discussion platform. In versions prior to 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1, a vulnerability in the discourse-subscriptions plugin allows users to gain access to subscription-gated groups without completing payment. This issue has been fixed in versions 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1.

discourse

Kitty is a cross-platform GPU based terminal. In versions 0.46.2 and below, the handle_compose_command() function in kitty/graphics.c performs bounds validation on composition offsets using unsigned 32-bit arithmetic that is subject to integer wrapping, potentially leading to Heap Buffer Over-Read/Write. An attacker who can write escape sequences to a kitty terminal (e.g., via a malicious file, SSH login banner, or piped content) can supply crafted x_offset/y_offset values that pass the bounds check after wrapping but cause massive out-of-bounds heap memory access in compose_rectangles(). No user interaction is required. No non-default configuration is required. The attacker only needs the ability to produce output in a kitty terminal window. This issue has been fixed in version 0.47.0.

kovidgoyal

Kitty is a cross-platform GPU based terminal. Versions 0.46.2 and below contain a heap buffer overflow in load_image_data() that allows any process which can write to the terminal's stdin to crash kitty immediately. The vulnerability is triggered by a single APC graphics protocol command with a PNG format declaration (f=100) whose payload exceeds twice the initial buffer capacity. The overflow is attacker-controlled in both length and content, causing DoS and potentially escalation to RCE itself. This issue has been fixed in version 0.47.0.

kovidgoyal

NanoMQ MQTT Broker (NanoMQ) is an all-around Edge Messaging Platform. In versions 0.24.10 and below, when NanoMQ handles high-concurrency reconnect traffic using a reconnect-collision payload, the broker can crash due to a NULL pointer dereference during MQTT session resumption for clean_start=0 clients. The transport's p_peer callback (tcptran_pipe_peer()) iterates cpipe->subinfol while copying session metadata from the cached old pipe to the new reconnecting pipe, without checking whether the pointer is NULL. Under a reconnect race, cpipe->subinfol can be freed and set to NULL before session restore invokes this function, resulting in a remote unauthenticated Denial-of-Service (process crash) condition. This issue has been fixed in version 0.24.11.

In the Linux kernel, the following vulnerability has been resolved: net: qrtr: ns: Limit the maximum server registration per node Current code does no bound checking on the number of servers added per node. A malicious client can flood NEW_SERVER messages and exhaust memory. Fix this issue by limiting the maximum number of server registrations to 256 per node. If the NEW_SERVER message is received for an old port, then don't restrict it as it will get replaced. While at it, also rate limit the error messages in the failure path of qrtr_ns_worker(). Note that the limit of 256 is chosen based on the current platform requirements. If requirement changes in the future, this limit can be increased.

Discourse is an open-source discussion platform. In versions prior to 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1, an authenticated user on a Discourse instance with the form templates feature enabled can read the name and structured content of form templates that are intended exclusively for categories they are not authorized to access. Impact is limited to disclosure of site configuration metadata. This issue has been fixed in versions 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1.

discourse

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. In versions 0.1.0 through 0.6.51, SendEmailBlock in autogpt_platform/backend/backend/blocks/email_block.py accepts a user-supplied smtp_server (string) and smtp_port (integer) as per-execution block inputs, then passes them directly to Python's smtplib.SMTP() to open a raw TCP connection with no IP address validation. This completely bypasses the platform's hardened SSRF protections in backend/util/request.py — the validate_url_host() function and BLOCKED_IP_NETWORKS blocklist that every other block uses to block connections to private, loopback, link-local, and cloud metadata addresses. An authenticated user on a shared AutoGPT deployment can use this to perform non-blind internal network port scanning and service fingerprinting: smtplib reads the target's TCP banner on connect and embeds it in the exception message, which is persisted as user-visible block output via the execution framework. This issue has been fixed in version 0.6.52.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. In versions 0.6.34 through 0.6.51, the backend deserializes Redis cache bytes using pickle.loads without integrity/authenticity checks. The write path serializes values with pickle.dumps(...) into Redis and the read path blindly invokes pickle.loads(...) on bytes with no HMAC/signature or strict schema validation gating deserialization. If an attacker can poison a shared-cache key in Redis, arbitrary command execution is possible in the backend container context, affecting confidentiality, integrity, and availability. This issue has been fixed in version 0.6.52.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Versions 0.4.2 through 0.6.51 are vulnerable to an unauthenticated Denial of Service (DoS) through the server due to uncontrolled disk space consumption. The download_agent_file endpoint creates persistent temporary files for every request but fails to delete them after they are served. An unauthenticated attacker can repeatedly call this endpoint to exhaust the server's disk space, causing the database or other system services to fail due to "No space left on device" errors, rendering the entire AutoGPT Platform backend unavailable to all users. This issue has been patched in version 0.6.52.

Discourse is an open-source discussion platform. In versions prior to 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1, outdated cached AI summaries can leak removed content to anonymous and unprivileged users who cannot regenerate summaries. This issue has been fixed in versions 2026.1.4, 2026.3.1, 2026.4.1 and 2026.5.0-latest.1. To work around this issue, restrict summary generation by tightening the allowed groups on the summarization Personas.

discourse

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Versions 0.6.36 through 0.6.50 are vulnerable to Authenticated Session Hijacking via IDOR. If an authenticated attacker can determine the session_id of another user's session, they can take it over, reading any messages in it and locking the legitimate user out. The PATCH /sessions/{session_id}/assign-user endpoint authenticates the caller but never verifies session ownership: the service layer invokes the session lookup with user_id=None, which the data access layer interprets as a privileged/system call that bypasses the ownership filter, allowing any authenticated user to reassign an arbitrary session to themselves. This issue has been patched in version 0.6.51.

Dokploy is a free, self-hostable Platform as a Service (PaaS). Versions 0.26.6 and below have OS command injection through the appName parameter. 3 chained issues cause this problem: inadequate input sanitization, lack of schema validation and direct shell interpolation. User-controlled application names are passed through inadequate sanitization (cleanAppName function only replaces spaces and converts to lowercase) before being interpolated directly into shell commands executed via execAsync() and execAsyncRemote(). An authenticated attacker can inject shell metacharacters (e.g., ;, $(), backticks, |, &) in the appName field during application creation, which are then executed with server-level privileges when service operations (start, stop, remove, scale) are triggered. This issue has been resolved in version 0.26.7.

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.8.0, GET /api/v1/memories/ef is accessible without authentication and executes request.app.state.EMBEDDING_FUNCTION(...). This allows any unauthenticated caller to trigger embedding generation which can lead to direct cost exposure if a paid provider is used. This vulnerability is fixed in 0.8.0.

openwebui

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.8.11, the API /api/v1/notes/{note_id} endpoint lacks proper authorization checks, allowing authenticated users to retrieve notes belonging to other users by guessing or enumerating UUIDs. This results in unauthorized disclosure of potentially sensitive or private user data. This vulnerability is fixed in 0.8.11.

openwebui

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.8.0, a Stored Cross-Site Scripting (XSS) vulnerability exists in the Banner component due to an improper sanitization order (specifically, DOMPurify is executed before the marked library). This vulnerability allows a compromised or malicious administrator to plant a malicious payload in the global banner. Crucially, this vector enables Privilege Escalation, as the malicious banner is rendered for all users, including the Super Admin (Primary Admin). Consequently, the payload successfully bypasses the existing security mechanism. An attacker can leverage this to steal the Super Admin's session token This vulnerability is fixed in 0.8.0.

openwebui