CVE-2024-3660-PoC

CVE-2024-3660 – 악성 모델을 통한 TensorFlow Keras 임의 코드 실행 개요 CVE-2024-3660 은 TensorFlow의 Keras 프레임워크에서 2.13 미만 버전에 영향을 주는 임의 코드 실행 취약점입니다. 공격자는 Lambda 레이어 등을 통해 모델 내부에 Python 코드를 삽입할 수 있으며, 이 코드는 모델이 로드될 때 자동으로 실행됩니다. 악성 .h5 모델 파일이 취약한 환경에서 로드되면, 로드 프로세스와 동일한 권한으로 임의 명령이 실행됩니다. 영향을 받는 버전 TensorFlow의...

CVE-2026-47117-openmed-rce

CVE-2026-47117: OpenMed PII 모델 로딩을 통한 인증되지 않은 원격 코드 실행 심각도: 치명적Critical, CVSS 4.0 9.3 , CVSS 3.1 9.8 CNA인 VulnCheck가 배정 벡터 v4.0: CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N 벡터 v3.1: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H 영향 범위: OpenMed 1.5.2 수정 버전: 1.5.2 CWE: CWE-94 코드 생...

CVE-2026-22807_Range

🛡️ CVE-2026-22807 시뮬레이션: AI 공급망 RCE 타깃 레인지 ⚠️ 면책 조항 / Disclaimer 이 프로젝트는 보안 연구 및 교육 목적 으로만 사용되며, AI 모델 로딩 과정에서의 공급망 공격 원리Remote Code Execution를 시연하기 위한 것입니다. 프로덕션 환경에서 생성된 악성 모델을 사용하지 마시고, 이 프로젝트를 불법 공격 행위에 사용하는 것을 엄격히 금지합니다. 개발자는 부적절한 사용으로 인한 법적 책임을 지지 않습니다. 📖 프로젝트 소개 Introduction 이것은...

CVE-2026-0596-Reproduction

CVE-2026-0596: MLflow 에코시스템의 안전하지 않은 역직렬화를 통한 임의 코드 실행 mlflow==2.11.1 및 mlserver==1.3.5 내 신뢰할 수 없는 모델 로딩 파이프라인과 관련된 검증, 기본 메커니즘 및 아키텍처 취약점을 상세히 설명하는 포괄적인 보안 연구 보고서입니다. ⚠️ 취약점 인텔리전스 권고: CVE-2026-0596 항목| 세부 정보 ---|--- 취약점 ID| CVE-2026-0596 / GHSA-rvhj-8chj-8v3c 공통 취약점 열거CWE| CWE-78: OS 명령에 사용되는 특수...

CVE-2026-100308

A flaw was found in GluonTS. The model loading component fails to safely handle untrusted data during deserialization, which can allow an attacker to execute arbitrary commands with the privileges of the running process. An attacker can exploit this issue by convincing a user or system to load a...

CVE-2026-100308

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. To remediate this issue, use...

CVE-2026-100308 GluonTS arbitrary command execution during model deserialization

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. To remediate this issue, use...

CVE-2026-100308

The Amazon GluonTS library is vulnerable to arbitrary command execution in versions prior to 0.17.0 . The root cause is the deserialization of untrusted data within the model loading component, which may allow context-dependent attackers to execute OS commands with the privileges of the loading p...

CVE-2026-100308 GluonTS arbitrary command execution during model deserialization

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. To remediate this issue, use...

CVE-2026-100308 GluonTS arbitrary command execution during model deserialization

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. To remediate this issue, use...

CVE-2025-32434

CVE-2025-32434: PyTorch RCE 취약점 - PoC CVE-2025-32434란? CVE-2025-32434 는 딥러닝 및 신경망에 널리 사용되는 Python 라이브러리인 PyTorch의 치명적인 원격 코드 실행RCE 취약점입니다. 핵심 문제 이 취약점은 안전한 것으로 알려진 weightsonly=True 매개변수를 사용하더라도 torch.load 함수에 존재합니다. weightsonly=True 옵션은 모델 파라미터만 로드하도록 제한하여 코드 실행을 방지하기 위해 설계되었습니다. 그러나 연구자들은 이 보호를...

CVE-2026-100308: Deserialization of Untrusted Data

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. To remediate this issue, use...

CVE-2026-93348 Unsloth Zoo Code Injection via model_type in config.json

Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the gettransformersmodeltype function in hfutils.py collects modeltype values from nested model configurations...

CVE-2026-93348 Unsloth Zoo Code Injection via model_type in config.json

Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the gettransformersmodeltype function in hfutils.py collects modeltype values from nested model configurations...

EUVD-2026-88318

Unsloth Zoo versions 2025.9.9 before 2026.8.14, as implemented in Unsloth 2025.9.9 through 2026.8.19, contains a code injection vulnerability in the model-loading compile path where the gettransformersmodeltype function in hfutils.py collects modeltype values from nested model configurations...

CVE-2026-54559

PocketSphinx (a lightweight speech recognizer, including its Python bindings) versions 5prealpha through 5.1.0 are affected by buffer overflow vulnerabilities in its model-loading code. The trie language-model loaders in src/lm/ngram_model_trie.c fail to validate boundary conditions in ARPA, DMP,...

CVE-2026-54559 PocketSphinx: Buffer overflows in language and acoustic model loading code

PocketSphinx is a small speech recognizer. Prior to 5.1.1, the trie language-model loaders in src/lm/ngrammodeltrie.c do not adequately validate boundary conditions in ARPA, DMP, and binary format headers, and the acoustic-model loaders in src/mdef.c and src/util/bio.c use sscanf with unbounded...

CVE-2026-54559 PocketSphinx: Buffer overflows in language and acoustic model loading code

PocketSphinx is a small speech recognizer. Prior to 5.1.1, the trie language-model loaders in src/lm/ngrammodeltrie.c do not adequately validate boundary conditions in ARPA, DMP, and binary format headers, and the acoustic-model loaders in src/mdef.c and src/util/bio.c use sscanf with unbounded...

CVE-2026-55832 Tract: Arbitrary file read via unsanitized ONNX external_data `location` (path traversal) on model load in tract-onnx

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.17, 0.22.3, and 0.23.2, the tract-onnx crate passes the attacker-controlled externaldata location from an ONNX model through onnx/src/tensor.rs getexternalresources and joins the value to the model...

CVE-2026-90553

A flaw was found in vLLM. The LlavaOnevision2 processor loader incorrectly ignores the trustremotecode parameter, which is designed to prevent the execution of untrusted code. This oversight allows an attacker to craft a malicious model containing arbitrary code. When such a model is loaded, the...