PULSE
EN VIVO0señales / 24h
FEED
vulnKEV agrega CVE-2026-21962 — Oracle / HTTP Server and Oracle Weblogic Server Proxy Plug-invulnKEV agrega CVE-2026-73570 — Synacor / Zimbra Collaboration Suite (ZCS)vulnKEV agrega CVE-2026-72530 — TrueConf / ServervulnKEV agrega CVE-2026-72529 — TrueConf / ServervulnKEV agrega CVE-2026-64849 — MLflow / MLflowvulnKEV agrega CVE-2026-33824 — Microsoft / Internet Key Exchange (IKE) Service ExtensionsvulnKEV agrega CVE-2026-59310 — Broadcom / VMware vCentervulnKEV agrega CVE-2026-55040 — Microsoft / SharePointvulnKEV agrega CVE-2026-65400 — Apple / macOSvulnKEV agrega CVE-2025-62593 — Ray-Project / Rayransomclop reclama a ZEBRA.COM · US · Manufacturingransomshinyhunters reclama a Metabase · US · Technologyransomshinyhunters reclama a Sharecare, Inc. · US · Healthcareransomthegentlemen reclama a IPS · IT · Not FoundvulnKEV agrega CVE-2026-21962 — Oracle / HTTP Server and Oracle Weblogic Server Proxy Plug-invulnKEV agrega CVE-2026-73570 — Synacor / Zimbra Collaboration Suite (ZCS)vulnKEV agrega CVE-2026-72530 — TrueConf / ServervulnKEV agrega CVE-2026-72529 — TrueConf / ServervulnKEV agrega CVE-2026-64849 — MLflow / MLflowvulnKEV agrega CVE-2026-33824 — Microsoft / Internet Key Exchange (IKE) Service ExtensionsvulnKEV agrega CVE-2026-59310 — Broadcom / VMware vCentervulnKEV agrega CVE-2026-55040 — Microsoft / SharePointvulnKEV agrega CVE-2026-65400 — Apple / macOSvulnKEV agrega CVE-2025-62593 — Ray-Project / Rayransomclop reclama a ZEBRA.COM · US · Manufacturingransomshinyhunters reclama a Metabase · US · Technologyransomshinyhunters reclama a Sharecare, Inc. · US · Healthcareransomthegentlemen reclama a IPS · IT · Not Found
← Todos los CVEs
CVE Watch24 ago 2026

CVE-2026-76841

Xinference loads models with Hugging Face remote code execution unconditionally enabled, and before version 2.12.0 exposes no setting to dis

CVSS

8.8

Alto

EPSS

KEV

Exploit Today

0

0-100

Publicado: 24 ago 2026 · Última mod.: 24 ago 2026 · CWE-94

EPSS · 30d

Sin historial EPSS suficiente todavía.

Descripción técnica

Xinference loads models with Hugging Face remote code execution unconditionally enabled, and before version 2.12.0 exposes no setting to disable it. Six loader call sites pass trust_remote_code=True as a literal or as an unconditional default: RerankModel._get_tokenizer in xinference/model/rerank/core.py, SentenceTransformerRerankModel.load in xinference/model/rerank/sentence_transformers/core.py, SentenceTransformerEmbeddingModel.load in xinference/model/embedding/sentence_transformers/core.py, FlagEmbeddingModel.load in xinference/model/embedding/flag/core.py, and two sites in xinference/model/llm/transformers/core.py where PytorchModel._sanitize_model_config and PytorchModel._get_components default the value to True. Because a caller with model launch access can register a model whose type is unknown and supply an arbitrary model path, the server reaches _auto_detect_type and then AutoTokenizer.from_pretrained, which imports and executes Python declared by the model directory's own tokenizer_config.json auto_map, running attacker-supplied code with the privileges of the worker process. Version 2.12.0 gates every site behind allow_trust_remote_code and the XINFERENCE_TRUST_REMOTE_CODE setting, permitting remote code only for bundled built-in models.

Referencias oficiales
CVEs relacionados
CVECVSSEPSSKEVRExplotTítuloVis.
CVE-2026-524909.8 CRÍ
0An issue in libtiff 85f2ac8e0b01cb7db2bbecf4a3b891bdbef67938 allows an attacker to execute arbitrary code via the process_command_opts() function in tools/tiffcrop.c8h
CVE-2026-408778.7 ALT
0Combodo iTop is a web-based IT service management tool. Prior to 3.2.3, iTop is vulnerable to PHP object injection in the user preference functionality, which can lead to remote code execution. This issue has been fixed in version 3.2.3.1d
CVE-2026-39975
0Combodo iTop is a web-based IT service management tool. Prior to 3.2.3, unauthenticated users could delete the .readonly file on iTop instances, leading to code execution. This file, created during the setup process, prevents users from performing write actions. This issue has been fixed in version 3.2.3.1d
CVE-2026-768368.8 ALT
0AzuraCast exposes the Liquidsoap custom configuration fields through an endpoint that does not require the permission guarding them. The backend_config property in backend/src/Entity/Station.php is annotated with GROUP_GENERAL, and PUT /api/station/{station_id}/profile/edit in backend/src/Controller/Api/Stations/ProfileEditController.php deserializes with that group while requiring only StationPermissions::Profile. AbstractArrayEntity::fromArray() then assigns every public property with no field-level permission check, so custom_config_top, custom_config, custom_config_pre_playlists, custom_config_pre_live, custom_config_pre_fade and custom_config_bottom are writable through it. ConfigWriter::writeCustomConfigurationSection() emits those values verbatim into the generated Liquidsoap .liq script, where the process.run() and process.exec() built-ins execute operating system commands when the backend restarts, which the built-in sync task triggers automatically once needs_restart is set. The dedicated endpoint for the same data, PUT /api/station/{id}/liquidsoap-config, requires StationPermissions::Broadcasting, so a station manager holding only the profile permission reaches configuration that the intended boundary reserves for broadcasting operators.1d
CVE-2026-783677.0 ALT
0A vulnerability was found in RPM's rpmbuild tarball processing. When processing a crafted source archive, the getTarSpec() function in tools/rpmbuild.cc passes an attacker-controlled tar archive member name to rpmExpand() as part of a %{basename:...} macro expression. A specially crafted .spec member name can therefore inject RPM macros, including Lua expressions, resulting in arbitrary code execution with the privileges of the user running rpmbuild. This can be exploited when a victim or automated build system processes an attacker-controlled source tarball using rpmbuild tarball mode (such as -ts, -ta, or -tb).1d
CVE-2026-781873.1 BAJ
34.1%
10A vulnerability has been found in Piwigo 16.3.0. This impacts an unknown function of the component Public Authentication Page. Such manipulation of the argument lang leads to cross site scripting. The attack may be performed from remote. A high complexity level is associated with this attack. The exploitability is said to be difficult. The exploit has been disclosed to the public and may be used. Upgrading to version 16.4.0 will fix this issue. The name of the patch is 5277a7dee4b8f1a174f1d69e1e2a4e1c82a3fc9e. It is recommended to upgrade the affected component.1d