high · CVSS v3 6.3 · CVSS v4 5.3 · EPSS 0.00316

CVE-2026-112086

ESPnet up to version 202610.post2 contains a deserialization flaw in the build_model function of inference_model.py. An attacker can manipul

Overview

ESPnet up to version 202610.post2 contains a deserialization flaw in the build_model function of inference_model.py. An attacker can manipulate the _target_ argument to trigger arbitrary code execution remotely. A public patch (f217b24f06de55e332cab10231469d5890dbbe00) is available.

Description

A vulnerability was found in ESPnet up to 202610.post2. This affects the function build_model of the file espnet3/publication/inference_model.py of the component Inference Model Handler. The manipulation of the argument _target_ results in deserialization. It is possible to launch the attack remotely. The exploit has been made public and could be used. The patch is identified as f217b24f06de55e332cab10231469d5890dbbe00. It is best practice to apply a patch to resolve this issue.

Impact

The vulnerability compromises confidentiality, integrity, and availability of systems running the affected ESPnet inference models. Remote code execution could allow attackers to take full control of the host, exfiltrate data, or disrupt services. Defenders should consider the risk when deploying ESPnet in production environments.

Remediation

Apply the official patch f217b24f06de55e332cab10231469d5890dbbe00 or upgrade to ESPnet 202610.post3 or later. Ensure that the build_model function is not exposed to untrusted input and validate all deserialization data. Monitor logs for suspicious deserialization activity and enforce least privilege on the inference service.

Risk context

The vulnerability is rated high severity with a CVSS v3 score of 6.3 and a low EPSS of 0.00316, indicating a moderate likelihood of exploitation but a significant impact if triggered. Timely patching is recommended to mitigate potential remote code execution.

Affected products

  • ESPnet Inference Model Handler
  • ESPnet 202610.post2

Scores

Severity
high
CVSS v2
7.5
CVSS v3
6.3
CVSS v4
5.3
EPSS
0.00316

deserialization remote-code-execution ESPnet high-severity patch inference-model

← All CVEs

Related CVEs

high

CVE-2026-112107

A flaw has been found in ztxz16 fastllm up to 1.8.2. Affected by this issue is the function SafeTensorItem::CreateBuffer of the file src/model.cpp of the component Safetensors Weight Loading. This manipulation of the argument data_offsets causes buffer overflow. The attack can be initiated remotely. The exploit has been published and may be used. The project was informed of the problem early through an issue report but has not responded yet.

View analysis

high

CVE-2026-108121

CVE-2026-108121 is a high‑severity flaw in SME Server 11.0rc1‑215 that lets a privileged user inject Apache directives during domain creation. The injected directives corrupt httpd.conf, causing Apache to fail to start after reboot and leading to a denial‑of‑service of the web server and its virtual hosts. The issue is exploitable only by users with high privileges on the host.

View analysis

high

CVE-2026-112081

A deserialization flaw in HanLP’s load function allows remote attackers to supply a crafted classpath argument, potentially leading to arbitrary code execution. The issue affects all versions up to 2.1.0-beta.64 and can be triggered via network interfaces. It has been publicly disclosed and remains unpatched.

View analysis

high

CVE-2026-112082

The CVE-2026-112082 vulnerability affects the SGLang library (up to version 0.5.19) in the Expert‑Distribution Recording Endpoint. It allows remote attackers to trigger a denial‑of‑service by manipulating the Scheduler.expert_distribution_handle function. The flaw is publicly documented and the fix is pending merge.

View analysis