Verifying Graceful Degradation in a Distributed Malware-Detection System with SPIN

Modern endpoint malware detection is distributed: a lightweight agent on each endpoint collects features from a scanned file or process, sends them to a remote server for analysis, and then enforces the returned verdict locally by blocking, quarantining, or disinfecting. Because the endpoint acts...

SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents

Language-model agents increasingly use tools to act on external systems. Earlier actions can alter files, permissions, database records, or other state, making a later routine-looking action harmful. Yet the visible interaction may not reveal the underlying state needed to assess that action. We...

JEV As a Judge for Agent Trace Security: An Empirical Comparison with Generative LLM Judges

Security evaluation of tool-using agents requires judging actions in context, yet generative judges add latency, explanation overhead, and output-validation failures. We study whether JEV, a typed decision model, offers a useful alternative for retrospective trace classification. We evaluate JEV...

FinRT: Distilling Adaptive Red-Teaming Strategies into Reusable Adversarial Generators in Consumer Finance

In regulated industries like consumer finance, seemingly harmless user queries can exploit large language model vulnerabilities, triggering safety failures and pushing responses dangerously close to policy limits. Existing automated red-teaming methods trade off attack effectiveness against...

PT-2026-99916

Name of the Vulnerable Software and Affected Versions Unsloth Zoo versions 2025.9.9 through 2026.8.13 Unsloth versions 2025.9.9 through 2026.8.19 Description A code injection issue exists in the model-loading compile path. The get transformers model type function in hf utils.py collects model typ...

API Secrets Should Never Become Tokens in the LLM's Vocabulary

Tool-using large language model LLM agents turn credential hygiene from a storage problem into an execution-security problem. A key pasted into a prompt, or embedded in a system prompt or tool configuration, crosses from an authentication boundary into a data pipeline, where it may persist in...

[SECURITY] [DSA 6517-1] nodejs security update

------------------------------------------------------------------------- Debian Security Advisory DSA-6517-1 [email protected] https://www.debian.org/security/ Moritz Muehlenhoff September 26, 2026 https://www.debian.org/security/faq -...

Use of Less Trusted Source

Overview vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs Affected versions of this package are vulnerable to Use of Less Trusted Source via incomplete propagation of revision and coderevision pins across artifact boundaries in the model loader subsystem. Wh...

DEBIAN-CVE-2026-100690

Hugo versions from v0.161.0 through v0.165.0 run Node.js tools css.PostCSS, css.TailwindCSS, js.Babel under the Node.js permission model to restrict file system reads to the project directory and configured mounts. Because the Node.js permission model validates only the lexical path and follows...

CVE-2026-100653

A flaw was found in vLLM. When operators configure a deployment to use a specific, pinned model revision, vLLM fails to apply the revision pin to certain supporting artifacts, such as model processors and configuration files. Consequently, these components are loaded from the external repository'...

CVE-2026-100690

Hugo versions v0.161.0 through v0.165.0 are vulnerable to an arbitrary file read due to an oversight in the Node.js permission model. While the model restricts file system reads to the project directory, it validates only the lexical path and follows symbolic links that point outside the allowed ...

EUVD-2026-87663

Hugo versions from v0.161.0 through v0.165.0 run Node.js tools css.PostCSS, css.TailwindCSS, js.Babel under the Node.js permission model to restrict file system reads to the project directory and configured mounts. Because the Node.js permission model validates only the lexical path and follows...

Insertion of Sensitive Information into Log File

Overview openclaw is a 🦞 OpenClaw — Personal AI Assistant Affected versions of this package are vulnerable to Insertion of Sensitive Information into Log File via the OpenAI-compatible transport layer, when a pinned third-party provider session whose resolved model metadata lacks a concrete base...

CVE-2026-100556

OpenClaw npm package openclaw versions = 2026.5.2 and command to reset the shared group session and persist a provider and model override. This allows a command-denied group member to select a provider and model already permitted by the operator for subsequent turns in the shared group session,...

CVE-2026-100545

The OpenClaw (npm package openclaw) before version 2026.8.1 contains a vulnerability where it incorrectly enforces sender tool policies during session-memory filename generation . In affected versions, the generation process creates an embedded helper that retains tools even if the originating se...

CVE-2026-100540 OpenClaw Feishu before 2026.8.1 Authentication Bypass via Disabled Account

OpenClaw Feishu before 2026.8.1 fails to validate whether a configured default account is disabled before selecting it for model tool operations. Attackers can exploit multi-account setups where a disabled default account retains credentials to read or modify Feishu resources through a revoked...

CVE-2026-100540 OpenClaw Feishu before 2026.8.1 Authentication Bypass via Disabled Account

OpenClaw Feishu before 2026.8.1 fails to validate whether a configured default account is disabled before selecting it for model tool operations. Attackers can exploit multi-account setups where a disabled default account retains credentials to read or modify Feishu resources through a revoked...

CVE-2026-100528

The OpenClaw npm package (openclaw ) versions prior to 2026.8.1 are vulnerable to a credential disclosure flaw. When using a third-party provider with an OpenAI-compatible API that lacks a concrete base URL in its metadata, a pinned session following a model configuration hot reload may retain th...

CVE-2026-100528

OpenClaw npm package 'openclaw' before 2026.8.1 could send third-party provider credentials to the wrong endpoint. In affected versions, when a third-party provider uses an OpenAI-compatible API and the resolved model metadata lacks a concrete base URL, a pinned session that continues after a mod...

CVE-2026-100528 OpenClaw before 2026.8.1 Credential Disclosure via Provider Endpoint

OpenClaw npm package 'openclaw' before 2026.8.1 could send third-party provider credentials to the wrong endpoint. In affected versions, when a third-party provider uses an OpenAI-compatible API and the resolved model metadata lacks a concrete base URL, a pinned session that continues after a mod...