CVEs

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, a caller can use the request-level media_io_kwargs field to select the GLMGA video backend and supply large values for the fps and max_frames options without a strict work ceiling. GLMGA constructs and deduplicates an attacker-sized pre-decode frame-index list, allowing a compact request and tiny valid video to consume disproportionate CPU time and memory in the shared media-loading executor. This issue is fixed

The WordPress Captcha Plugin by Captcha Bank plugin for WordPress is vulnerable to Reflected Cross-Site Scripting due to the use of add_query_arg without appropriate escaping on the URL in all versions up to, and including, 4.0.36. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that execute if they can successfully trick a user into performing an action such as clicking on a link.

A flaw was found in libtirpc. An authenticated remote attacker could exploit this vulnerability by sending oversized wrapped data requests to an RPC (Remote Procedure Call) service configured with RPCSEC_GSS (Remote Procedure Call Security-Generic Security Service) integrity or privacy. This can lead to unbounded memory allocation, resulting in a Denial of Service (DoS) by exhausting system resources and causing the service to become unstable or terminate.

Tandoor Recipes is an application for managing recipes, planning meals, and building shopping lists. Prior to 2.6.10, executes user-controlled Automation regular-expression patterns with Python re.sub() without complexity validation or an execution timeout when an authenticated user stores a catastrophic-backtracking regex and any user in the same Space imports a matching recipe. Request processing uses Automation.param_2, apply_regex_replace_automation, sub, api/recipe-from-source/. The attack

Improper verification of the source of a communication channel in the DNS resolver in Amazon FreeRTOS-Plus-TCP before V4.4.2 might allow remote third-party to insert a chosen address into the device's DNS cache via a forged DNS reply. To remediate this issue, users should upgrade to version V4.4.2 and enable ipconfigDNS_CHECK_REPLY_SOURCE_IP.

IBM i 7.3, 7.4, 7.5, and 7.6 could allow a local attacker to obtain sensitive information due to missing subtree checks during file handle resolution.

IBM i 7.3, 7.4, 7.5, and 7.6 is vulnerable to a denial of service to Digital Certificate Manager (DCM) for System TLS using applications attempting to create secure network connections. A bad actor could use this to prevent a service from starting or cause it to fallback to non-secure connections.

IBM Db2 Genius Hub 1.0 could allow a local attacker to obtain sensitive information due to improper verification of cryptographic signature.

IBM i 7.3, 7.4, 7.5, and 7.6 is could allow a remote authenticated attacker to bypass security restrictions due to improper enforcement of the ChrootDirectory configuration directive.

IBM i 7.3, 7.4, 7.5, and 7.6 could allow a remote authenticated attacker to bypass security restrictions due to improper authentication.

IBM i 7.3, 7.4, 7.5, and 7.6 could allow a local attacker to bypass NFS version restrictions due to improper enforcement of the export version flag.

Openwave Webtop Media Transcoder does not enforce administrator-defined file type restrictions to restrict the file-types to be converted. This may allow an authenticated user to process and convert unintended document types if bypassing the Webmail experience.

Antiword through 0.37 contains a heap out-of-bounds read vulnerability in vGet8LstInfo in prop8.c caused by an inverted bounds check on the PlcfLst record count. Attackers can supply a crafted Word 8 .doc file claiming up to 0xffff LSTF records in a tiny buffer, causing reads past the heap allocation that crash the process.

Cube-studio through 2026.09.01 contains a missing authorization vulnerability in myapp/views/view_pipeline.py that allows authenticated users to view or copy other project groups' pipelines by id. Attackers can enumerate sequential pipeline ids via the web, log, monitoring, pod, and copy_pipeline endpoints to expose DAGs, cluster names, namespaces, Argo workflow ids, and pod names, or clone pipelines.

REBUILD through 4.5.0 contains a missing authorization vulnerability in TaskCommentController.commentList that allows authenticated low-privileged users to read task comments without project membership checks. Attackers can supply the id of a task in a private project to the /project/comments/list endpoint to retrieve its comment content and attachment references.

Aureus ERP through 1.6.0 contains an authorization bypass vulnerability that allows authenticated API users to create or move records into disallowed companies by supplying a foreign company_id. Attackers can submit create or update requests to plugin REST API endpoints such as partners, sales, or inventories to break multi-company data segregation.

Allegro Ralph through 20260902.1 contains a missing authorization vulnerability that allows authenticated staff users to bypass rack model permissions because RalphPermission skips checks for views without a queryset. Attackers can call DCAssetsView and SRRacksAPIView at /api/rack/<id> and /api/server_room/<id> to read asset hostnames, serials and management IPs, and to modify, move or create racks.

Exrick XBoot through 3.3.4 contains an authorization bypass vulnerability that allows authenticated users to evade role data-scope restrictions by sending openDataFilter=false. Attackers can append this parameter to the department getByParentId or search endpoints to retrieve the entire organization's department tree and names outside their assigned scope.