Vulnerabilities

llama.cpp builds through commit 97f06e9 fail to handle a negative top_n value in POST requests to the /rerank endpoint, producing an out-of-memory condition (std::bad_alloc, classified as uncontrolled recursion per CWE-674) that is surfaced as an HTTP 500 error. The flaw is triggered remotely and without authentication whenever a server process is started with the --reranking flag and accepts requests at /rerank. An attacker who can reach the endpoint can disrupt availability, either crashing the process or making rerank requests fail, with no confidentiality or integrity impact (CVSS 3.1: AV:N/AC:L/PR:N/UI:N, A:H). Only deployments running the llama.cpp HTTP server with reranking enabled are affected; instances run without the --reranking flag, or whose /rerank endpoint is not reachable by untrusted clients, are not exposed. No public proof of concept exists, it is not in CISA's KEV catalog, and EPSS estimates a 0.3% probability of exploitation within 30 days, so no in-the-wild exploitation is currently known.

Do: Upgrade to a llama.cpp build newer than commit 97f06e9 once a patched release is available, since the advisory does not name a fixed version. As interim mitigations, avoid starting the server with --reranking unless reranking is required, restrict /rerank to trusted clients or place authentication in front of it, and reject requests with a negative top_n at a reverse proxy. Monitor /rerank for HTTP 500 responses as a possible sign of probing.

llama.cpp builds b1886 through b7445 contain a double free vulnerability in the LLaMA-Android JNI wrapper where new_1batch() allocates memory using malloc() while free_1batch() deallocates it using the C++ delete operator, causing heap metadata corruption. Attackers can trigger this memory management mismatch to cause denial of service through process crashes or potentially achieve arbitrary code execution depending on allocator state.

Python StateMachine versions 3.0.0 before 3.2.0 contains a remote code execution vulnerability that allows attackers to execute arbitrary code by supplying malicious SCXML documents containing crafted ` ` attributes evaluated unsafely. The SCXMLProcessor passes attacker-controlled expression strings through a call chain ending in Python's built-in eval() without sandboxing, enabling arbitrary code execution in the context of the hosting process.

Gotenberg is a Docker-powered stateless API for PDF files. Prior to 8.31.0, Gotenberg's /forms/pdfengines/metadata/write HTTP endpoint accepts a JSON metadata object and passes its keys directly to ExifTool via the go-exiftool library. No validation is performed on key characters. A \n embedded in a JSON key splits the ExifTool stdin stream into a new argument line, allowing an attacker to inject arbitrary ExifTool flags — including -if, which evaluates Perl expressions. This achieves unauthenticated OS command execution in a single HTTP request. The response is HTTP 200 with a valid PDF, making the attack transparent to basic monitoring. This vulnerability is fixed in 8.31.0.

Gotenberg is a Docker-powered stateless API for PDF files. In versions 8.30.1 and earlier, the metadata write endpoint validates metadata keys for control characters but leaves metadata values unsanitized. A newline character in a metadata value splits the ExifTool stdin line into two separate arguments, allowing injection of arbitrary ExifTool pseudo-tags such as -FileName, -Directory, -SymLink, and -HardLink. This is a bypass of the incomplete key-sanitization fix introduced in v8.30.1. An unauthenticated attacker can rename or move any PDF being processed to an arbitrary path in the container filesystem, overwrite arbitrary files, or create symlinks and hard links at arbitrary paths.

Gotenberg is an API-based document conversion tool. In versions 8.30.1 and earlier, the default private-IP deny-lists for the --webhook-deny-list and --api-download-from-deny-list flags use a case-sensitive regular expression (^https?://) to match URL schemes. Because Go's net/url.Parse() normalizes the scheme to lowercase before establishing the outbound TCP connection, an attacker can bypass the deny-list by simply capitalizing part of the URL scheme (e.g., HTTP://, HTTPS://, or Http://). This allows unauthenticated requests to reach internal network services, including private IP ranges, loopback addresses, and cloud instance metadata endpoints such as HTTP://169.254.169.254/latest/meta-data/. This bypasses the same security control that was patched in CVE-2026-27018. This issue has been fixed in version 8.31.0.

A vulnerability in the web application allows standard users to escalate their privileges to those of a super administrator through parameter manipulation, enabling them to access and modify sensitive information.

Gotenberg is an API for converting document formats. In 8.29.1 and earlier, Gotenberg uses dlclark/regexp2 to compile user-supplied scope patterns without setting a proper timeout. Users with access to features using this logic can hang workers indefinitely.

llama.cpp is an inference of several LLM models in C/C++. Prior to version b8492, the RPC backend's deserialize_tensor() skips all bounds validation when a tensor's buffer field is 0. An unauthenticated attacker can read and write arbitrary process memory via crafted GRAPH_COMPUTE messages. Combined with pointer leaks from ALLOC_BUFFER/BUFFER_GET_BASE, this gives full ASLR bypass and remote code execution. No authentication required, just TCP access to the RPC server port. This issue has been patched in version b8492.

Gotenberg is an API for converting document formats. Prior to version 8.29.0, the fix introduced for CVE-2024-21527 can be bypassed using mixed-case or uppercase URL schemes. This issue has been patched in version 8.29.0.

A Reflected Cross-Site Scripting (XSS) vulnerability exists in SourceCodester Sales and Inventory System 1.0. The vulnerability is located in the add_stock.php file via the "msg" parameter. The application fails to sanitize the input, allowing remote attackers to inject arbitrary web script or HTML via a crafted URL.

BlueKitchen BTstack versions prior to 1.8.1 contain an out-of-bounds read vulnerability in the AVRCP Controller GET_PLAYER_APPLICATION_SETTING_ATTRIBUTE_TEXT and GET_PLAYER_APPLICATION_SETTING_VALUE_TEXT handlers that allows nearby attackers to read beyond packet boundaries. Attackers can establish a paired Bluetooth Classic connection and send specially crafted VENDOR_DEPENDENT responses to trigger out-of-bounds reads, causing information disclosure and potential crashes on affected devices.

A Reflected Cross-Site Scripting (XSS) vulnerability exists in SourceCodester Sales and Inventory System 1.0 in the view_product.php file via the "limit" parameter. The application fails to sanitize the input, allowing remote attackers to inject arbitrary web script or HTML via a crafted URL.

A vulnerability was determined in SourceCodester Sales and Inventory System 1.0. This vulnerability affects unknown code of the file /update_stock.php of the component HTTP GET Parameter Handler. This manipulation of the argument sid causes sql injection. Remote exploitation of the attack is possible. The exploit has been publicly disclosed and may be utilized.

A vulnerability was found in SourceCodester Sales and Inventory System 1.0. This affects an unknown part of the file /update_sales.php of the component HTTP GET Parameter Handler. The manipulation of the argument sid results in sql injection. The attack may be launched remotely. The exploit has been made public and could be used.

A security vulnerability has been detected in SourceCodester Sales and Inventory System 1.0. This issue affects some unknown processing of the file update_customer_details.php of the component HTTP GET Parameter Handler. Such manipulation of the argument sid leads to sql injection. The attack can be executed remotely. The exploit has been disclosed publicly and may be used.

llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.