one-multiply-too-many-cve-2026-70638-llama-cpp-android-jni-integer-overflow

CVE-2026-70638 - Estouro de Inteiro no JNI LLaMA-Android do llama.cpp PoC complementar para o artigo da Hunt-Benito "Uma Multiplicação a Mais: CVE-2026-70638 — Estouro de Inteiro na Alocação de Heap do JNI Android do llama.cpp" →...

LlamaStack-RCE-Deterministic-Supply-Chain-Exploitation-Hardening-Framework-CVE-2024-50050-

= doi.org/10.5281/zenodo.18278567 = orcid.org/0009-0007-7728-256X 💀 LlamaStack-RCE: CVE-2024-50050 Framework di Sfruttamento della Supply Chain MATERIALE CLASSIFICATO SENZA RESTRIZIONI - SOLO PER RED TEAMING AUTORIZZATO 1. Sintesi Esecutiva Analisi Deterministica LlamaStack-RCE è un framework di...

Llama-Stack-0.4.0rc3-local-CVE-2026-25211

PoC per la Perdita della Password pgvector di Llama Stack CVE-2026-25211 Proof-of-Concept Locale che dimostra l'esposizione in chiaro della password del database nei log di inizializzazione delle versioni 0.4.0rc3 di llama-stack. Panoramica Nelle versioni interessate di llama-stack precedenti all...

AI-Vulnerabilities-Playground

AIVP Playground für KI-Schwachstellen AIVP ist eine praxisnahe Plattform für KI-Sicherheits-Labs, die 4 Phasen und über 50 Labs umfasst derzeit über 50 Szenarien in der Codebasis. Sie kombiniert Ollama + Llama 3.1 , ein FastAPI -Backend und ein React/Vite -Frontend, damit Praktiker offensive und...

llm-censorship-steering

LLM Censorship Steering Questo repository contiene l'implementazione del codice per Steering the CensorShip: Uncovering Representation Vectors for LLM "Thought" Control di Hannah Cyberey e David Evans. Presentiamo un metodo che individua i "vettori di steering" nei meccanismi interni degli LLM pe...

llama-factory-webui-rce-cve-2026-58116-trust-remote-code-model-path-injection

CVE-2026-58116 — RCE nella WebUI di LLaMA-Factory tramite trustremotecode Prova di concetto per CVE-2026-58116 CVSS 9.8 Critico: esecuzione remota di codice in LLaMA-Factory = v0.9.5. La WebUI di LLaMA-Factory imposta in modo hardcoded trustremotecode=True quando carica un modello...

D-Scan

Quando il Contesto Morde: Rilevare l'Avvelenamento RAG tramite il Collasso dell'Attenzione a Livello di Documento D-SCAN è un framework di analisi per rilevare attacchi di avvelenamento dei documenti nei sistemi RAG Retrieval-Augmented Generation. Raccoglie gli stati interni dell'LLM durante la...

llama_facts

Fatti sui lama Questo progetto è stato originariamente creato per il concorso CTF Women in Cybersecurity WIYCS 2022 del Rochester Institute of Technology RIT. È vulnerabile alla Server-Side Template Injection SSTI, definita nella controversa CVE CVE-2019-8341. Descrizione della sfida Descrizione:...

statecraftcyber

Statecraft Cyber Intelligence Plataforma de threat intelligence em português construída do zero como projeto pessoal de Blue Team. Agrega dados de fontes abertas globais e os transforma em briefings técnicos, notícias classificadas, CVEs enriquecidos e IOCs estruturados — tudo em PT-BR, atualizad...

Exponentiated-Gradient-Descent-LLM-Attack

Modifica il file Readme. Questo è un progetto che esplora il metodo di ottimizzazione Exponentiated Gradient Descent per produrre suffissi avversari con cui attaccare modelli linguistici di grandi dimensioni allineati. Il metodo si è dimostrato efficace sul modello chat Llama-2 con 7 miliardi di...

CVE-2026-43631

A flaw was found in llama.cpp. This use-after-free vulnerability in the llama-server component, specifically within the vocab pointer when the --sleep-idle-seconds feature is enabled, allows unauthenticated remote attackers to execute arbitrary code. Attackers can exploit this by sending requests...

SUSE CVE-2026-43632

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints /tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/counttokens that bypass the task queue and access ctxserver.vocab directly on HTTP...

EUVD-2026-54280

llama.cpp builds b1886 through b7445 contain a race condition use-after-free vulnerability in the LLaMA-Android JNI wrapper where bench1model and free1context lack synchronization, allowing Thread A to operate on freed memory while Thread B concurrently frees the llamacontext. Attackers can explo...

EUVD-2026-54276

llama.cpp builds b1886 through b7445 contain an integer overflow vulnerability in the LLaMA-Android JNI wrapper where the new1batch function multiplies sizeofllamaseqid by an attacker-controlled nseqmax parameter without overflow validation, causing heap buffer allocation to wrap and allocate...

EUVD-2026-54287

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints /tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/counttokens that bypass the task queue and access ctxserver.vocab directly on HTTP...

EUVD-2026-54286

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in the vocab pointer of llama-server when the --sleep-idle-seconds feature is enabled, allowing unauthenticated remote attackers to execute arbitrary code. Attackers can trigger the vulnerability by sending...

Linux Distros Unpatched Vulnerability : CVE-2026-43631

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in the vocab pointer of llama-server when the --sleep-idle-seconds featu...

Linux Distros Unpatched Vulnerability : CVE-2026-43629

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - llama.cpp builds b4882 through b9058 contain a heap buffer overflow vulnerability in the KV cache state restore path where the statereaddata function computes...

Linux Distros Unpatched Vulnerability : CVE-2026-43628

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - llama.cpp builds b3978 through b9058 contain an integer underflow and out-of-bounds read vulnerability in the DRY sampler that allows unauthenticated attackers ...

Linux Distros Unpatched Vulnerability : CVE-2026-43632

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints /tokenize,...