Threat database

vLLM from 0.7.3 before 0.31.0 contains an uncontrolled resource consumption vulnerability that allows remote clients to control multimodal processor parameters by supplying per-request mm_processor_kwargs. Attackers can set large image size, patch count, crop count, or frame count values in Isaac, InternVL, H2OVL, and other processors to exhaust memory and terminate the shared server.

vLLM versions before 0.31.0 have an improper access control vulnerability involving cache aliasing. Authenticated attackers can manipulate shared cache keys by supplying controlled values like EXIF ImageID or cache_salt. This allows attackers to access cached processed media from other requests or reuse key-value state, potentially leading to unauthorized data access.

vLLM versions from 0.7.3 up to but not including 0.31.0 contain a code injection vulnerability. This occurs because the software forwards user-supplied code revision parameters to Transformers AutoProcessor.from_pretrained when trust_remote_code is enabled, allowing an attacker to execute arbitrary Python code via the unauthenticated /tokenize endpoint.

vLLM after 0.24.0 and before 0.31.0 contains a memory exhaustion vulnerability in MOSS-Audio model serving that caches a processor for every distinct request-controlled mm_processor_kwargs value in an unbounded dictionary. Unauthenticated attackers can send distinct or oversized mel_config values to the /tokenize endpoint, which bypasses --api-key, to exhaust host memory and crash the API process.

vLLM versions before 0.31.0 have a vulnerability where the min_tokens <= max_tokens check is skipped if max_tokens is null in /inference/v1/generate requests. This causes an uncaught exception that hangs all subsequent generation requests until the service is manually restarted. The vulnerability can be triggered by unauthenticated remote attackers sending a single request with an oversized min_tokens value.

vLLM before 0.31.0 uses a client-supplied multimodal uuid verbatim as cache identity with no consistency check on cache hits, allowing engine crashes. Unauthenticated attackers can desynchronize the P0/P1 processor caches to trigger a fatal token-placeholder mismatch, killing the entire serving instance until restarted.

vLLM versions before 0.31.0 have a vulnerability in the InputBatch.condense() function that causes a cross-request mask leak. This flaw allows remote attackers to influence other clients' output by leaving stale allowed_token_ids rows, which results in token allowlists from one request being applied to another. Attackers can exploit this by submitting concurrent constrained generate requests, including through unauthenticated /invocations endpoints.

WWBN AVideo versions before 24.0 contain a cross-site request forgery (CSRF) vulnerability in legacy endpoints such as playlistStatus.php. This vulnerability bypasses the autoCSRFGuard() check that applies only to .json.php endpoints. Exploitation can allow attackers to trick logged-in users into performing unauthorized actions like making private playlists public, removing videos from playlists, or forging video suggestions.

Astral uv before 0.8.22 contains a path traversal vulnerability in source distribution (sdist) tar extraction that allows a specially crafted sequence of symlinks to redirect file writes outside the intended extraction prefix. Attackers can publish a malicious sdist that, when installed by a victim, writes arbitrary files to filesystem locations outside the installation prefix.

An improper access control vulnerability in the Ozone component of Google Chrome prior to version 156.0.8078.47 allows a remote attacker to leak sensitive information via a crafted HTML page. The vulnerability has a medium severity rating and does not impact integrity or availability. Exploitation requires user interaction.

An information leak vulnerability exists in ANGLE component of Google Chrome on Windows versions prior to 156.0.8078.47. This flaw allows a remote attacker to leak sensitive information by convincing a user to visit a crafted HTML page. The vulnerability has a medium severity rating with a CVSS score of 6.5.

An improper access control vulnerability in DevTools of Google Chrome versions prior to 156.0.8078.47 could allow a remote attacker to leak sensitive information by tricking a user into visiting a crafted HTML page. The vulnerability has a medium severity rating and does not impact integrity or availability.

A medium severity vulnerability (CVE-2026-113129) in Google Chrome extensions prior to version 156.0.8078.47 allows a remote attacker to potentially leak sensitive information by exploiting improper access control via a crafted HTML page. The vulnerability does not impact integrity or availability but can lead to confidentiality loss.

CVE-2026-113738 is a path traversal vulnerability in the extract_mtp_heads function of the qwen_mtp_gguf_pipeline.py script within the MTP weight_map Handler component of R6410418 Jackrong-llm-finetuning-guide. The vulnerability allows remote attackers to manipulate the shard_name argument to perform path traversal. The product uses a rolling release model, so no specific affected or fixed versions are provided. The issue was reported early but the project has not responded yet. The CVSS 4.0 score is 5.3, indicating medium severity.

CVE-2026-114038 is a medium severity vulnerability in Red Hat Enterprise Linux 10's device-mapper-multipath. A local user can cause an out-of-bounds read by sending a specially crafted command with a format string ending in a trailing percent character to the multipath daemon. This can lead to information disclosure or cause the daemon to crash, resulting in denial of service.

CVE-2026-113737 is a path traversal vulnerability in the R6410418 Jackrong-llm-finetuning-guide project affecting the validate_indexed_tensors function in the qwen_mtp_gguf_pipeline.py script. Manipulation of the weight_map argument can lead to path traversal, potentially exploitable remotely. The project uses a rolling release system, so specific affected versions are not disclosed. The vulnerability has a medium severity with a CVSS 4.0 score of 5.1. The issue was reported early but the project has not yet responded or released a fix. Exploit code has been published but no known exploitation in the wild is reported.

CVE-2026-113736 is a critical vulnerability in the D-Link DHP-W610AV device version 1.02EUb03. It involves the login_prepare_hnap function in the HNAP Login component, where manipulation of the Password argument results in an empty password being written to the configuration file. This vulnerability can be exploited remotely without authentication. The exploit has been publicly disclosed, increasing the risk of exploitation.

ProSoft RLX-IFH24E version 2.2.44 contains a vulnerability where an unsigned firmware upgrade package is executed with root privileges.

Parasoft Continuous Testing Platform version 10.7.5 is affected by a JSP command execution vulnerability.

An exploit targeting Exasol Database version 7.1.30 enables command execution. Limited technical details are available, and no affected versions or patch information are provided.