CVE-2026-105754

A flaw was found in vLLM. The scale-out inference service fails to properly validate caller-supplied multimodal parameters, such as image geometry and cache identifiers, against the active model's configuration. An authenticated remote attacker can exploit this vulnerability by submitting crafted...

BARE-AI: Bit-Flip Attack Resilience in AI Hardware through Built-In Performance Monitors

Deep Neural Networks DNNs are integral to many safety critical systems, yet they remain highly vulnerable to bit-flip attacks BFAs, where a few memory level perturbations can drastically degrade accuracy. Existing defenses incur significant hardware overhead, depend on retraining, or fail against...

EUVD-2026-92844

vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features...

CVE-2026-105754

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

gradient-untangler

gradient-untangler A local research harness that searches for the exact tokens that make an open-weight language model start its answer the way you specify. Open-weight models ship as ordinary files: config.json, a tokenizer, and one or more .safetensors or .bin shards. A Hugging Face id is only ...

CVE-2026-105755 vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induced errors on `/score` and `/rerank`

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's querykey value from the caller-controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can...

CVE-2026-105754 vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

CVE-2026-105754 vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

CVE-2026-47630

CVE-2026-47630 — NVIDIA Triton Inference Server: TRITONBATCHSTRATEGYPATH를 통한 임의 dlopen triton-inference-server/core의 사용자 정의 배치 전략 로더에서 발생하는 절대 경로 탐색Absolute Path Traversal 취약점입니다. 모델 구성 매개변수가 검증 없이 dlopenRTLDNOW | RTLDLOCAL에 전달되므로, 모델의 config.pbtxt에 영향을 줄 수 있는 사람은 누구나 모델 로드 시점에 Triton 서버 프로세스 내부에...

Astra Linux – Vulnerability in Firefox

While the text displayed in Autofill tooltips cannot be directly read by JavaScript, it was rendered using page fonts. Side-channel attacks on the text using specially crafted fonts could have led to this text being interpreted by the webpage. This vulnerability affects Firefox versions earlier...

EntropicaPublic

March updates I need to move some work over from the private repos. The exploration into using massivly multheaded models with mixer and readout functions separated continues. Currently training up a Chinese TinyStories on a 3B-equivalent model that trains, easily, on an 8GB gpu. We're looking at...

vicidial-cve-2024-8503-blind-sqli-poc

vicidial-cve-2024-8503-blind-sqli-poc VICIdial의 VERMAJAXfunctions.php 엔드포인트CVE-2024-8503에 대한 인증되지 않은 시간 기반 블라인드 SQL 인젝션 개념 증명PoC입니다. 이 프로젝트는 승인된 보안 테스트, 검증 및 교육 목적으로만 제공됩니다. 개요 CVE-2024-8503은 VICIdial에서 인증되지 않은 SQL 인젝션으로, 조작된 Basic Auth 사용자 이름 값이 VERMAJAXfunctions.php의 취약한 쿼리 로직에 도달할 수 있습니다. 이 Po...

aotopsy

AOTopsy A Dart AOT snapshot analyzer. Turns libapp.so — the compiled Dart code inside a Flutter release APK — into function names, class layouts, call graphs, behavioral signals, and readable pseudocode. No Dart VM, no SDK compilation, no runtime fallback. Fork notice: AOTopsy is a fork of...

CVE-2023-43654

CVE-2023-43654 ShellTorch 는 PyTorch의 기본 추론 서버인 PyTorch TorchServe에서 발견된 3개의 치명적 취약점 체인입니다. 이는 네트워크를 통해 임의의 원격 코드 실행으로 이어지는 SSRF 취약점에 대한 종단 간 POC입니다. 1. TorchServe 0.8.1 실행하기 docker run -d --rm -it -p 8080:8080 -p 8081:8081 -p 8082:8082 -p 7070:7070 -p 7071:7071 --name shelltorch-demo...

Hidden Risks of Jev: An Empirical Study of Security, Privacy, and Dual Use

Jev turns natural-language questions into typed answers and probabilities with low latency and cost, enabling applications to route requests and select tools. While this interface allows Jev to integrate naturally into application workflows as a decision layer, the security and privacy implicatio...

Static Bootstrap Placement for Encrypted Language Model Decoding

Language models increasingly serve prompts that carry private data, and secure inference under homomorphic encryption lets a client outsource the computation without revealing the prompt. Existing secure inference systems run a forward pass without consuming a token under encryption, and generati...

CVE-2026-101918 vulnerabilities

Vulnerabilities for packages: openstack-glance-2025.1-fips, superset-fips, openstack-horizon-2026.1-fips, openstack-manila-2025.2, nemo-rl-cuda-13.0, openstack-glance-2025.2, openstack-keystone-2025.1, openstack-cinder-2026.1, request-1276, keep-fips, openstack-keystone-2025.2-fips, awx,...

GHSA-42VR-XJ54-VC7V vulnerabilities

Vulnerabilities for packages: openstack-glance-2025.1-fips, superset-fips, openstack-horizon-2026.1-fips, openstack-manila-2025.2, nemo-rl-cuda-13.0, openstack-glance-2025.2, openstack-keystone-2025.1, openstack-cinder-2026.1, request-1276, keep-fips, openstack-keystone-2025.2-fips, awx,...

GHSA-HJ66-6F7G-4R5V vulnerabilities

Vulnerabilities for packages: pgadmin4-fips, kubeflow-jupyter-web-app, kserve-xgbserver, kserve-lgbserver-fips, ansible-operator, keep-fips, openstack-keystone-2025.2-fips, awx, authentik-fips, mlrun, k8s-sidecar, kubeflow-pipelines, kubeflow-volumes-web-app, azureml-inference-server-http-fips,...

GHSA-XPV3-W29H-X7CV vulnerabilities

Vulnerabilities for packages: pgadmin4-fips, kubeflow-jupyter-web-app, kserve-xgbserver, kserve-lgbserver-fips, ansible-operator, keep-fips, openstack-keystone-2025.2-fips, awx, authentik-fips, mlrun, k8s-sidecar, kubeflow-pipelines, kubeflow-volumes-web-app, azureml-inference-server-http-fips,...