AgentTrust: Runtime Safety Evaluation and Interception for AI Agent Tool Use

Modern AI agents execute real-world side effects through tool calls such as file operations, shell commands, HTTP requests, and database queries. A single unsafe action, including accidental deletion, credential exposure, or data exfiltration, can cause irreversible harm. Existing defenses are...

MOSAIC-Bench: Measuring Compositional Vulnerability Induction in Coding Agents

Coding agents often pass per-prompt safety review yet ship exploitable code when their tasks are decomposed into routine engineering tickets. The challenge is structural: existing safety alignment evaluates overt requests in isolation, leaving models blind to malicious end-states that emerge from...

CVE-2026-42086

OpenC3 COSMOS provides the functionality needed to send commands to and receive data from one or more embedded systems. Prior to version 7.0.0, the Command Sender UI uses an unsafe eval function on array-like command parameters, which allows a user-supplied payload to execute in the browser when...

CVE-2026-42079 PPTAgent: Arbitrary Code Execution via Python eval() of LLM-Generated Code with Builtins in Scope

PPTAgent is an agentic framework for reflective PowerPoint generation. Prior to commit 418491a, PPTAgent is vulnerable to arbitrary code execution via Python eval of LLM-generated code with builtins in scope. This issue has been patched via commit 418491a...

CLSA-2026-1777878328 libxml2: Fix of 2 CVEs

CVE-2018-14404: fix NULL pointer dereference in xmlXPathCompOpEval when parsing an invalid XPath expression in the XPATHOPAND or XPATHOPOR case - CVE-2019-19956: fix memory leak in xmlParseBalancedChunkMemoryRecover related to newDoc-oldNs...

A Validated Prompt Bank for Malicious Code Generation: Separating Executable Weapons from Security Knowledge in 1,554 Consensus-Labeled Prompts

Existing benchmarks of language-model refusal on malicious-coding tasks routinely conflate requests for executable malicious software with requests for harmful security knowledge. This conflation matters because the two request types plausibly trigger distinct refusal pathways in safety-aligned...

Observability for Post-Quantum TLS Readiness: A Multi-Surface Evidence Framework

Post-quantum migration in Transport Layer Security TLS requires evidence-aware measurements that distinguish session negotiation, endpoint capability, certificate-chain evidence, and the provenance of missing observations. This distinction is essential under TLS 1.3 encryption, resumption, mutual...

FIRCE: A Framework for Intrusion Response and Conformal Evaluation

Machine learning-based intrusion detection systems deployed in real-world environments frequently suffer from model degradation due to concept drift, where changes in traffic patterns invalidate training assumptions. To address this, we present FIRCE, a Framework for Intrusion Response and...

CVE-2026-2052 Widget Options <= 4.2.2 - Authenticated (Contributor+) Remote Code Execution via Display Logic

The Widget Options – Advanced Conditional Visibility for Gutenberg Blocks & Classic Widgets plugin for WordPress is vulnerable to Remote Code Execution in all versions up to, and including, 4.2.2 via the Display Logic feature. This is due to the plugin using eval on user-supplied Display Logic...

WordPress plugin Widget Options – Advanced Conditional Visibility for Gutenberg Blocks & Classic Widgets 代码注入漏洞

WordPress and WordPress plugins are both products of the WordPress Foundation. WordPress is a blog platform developed using the PHP language. This platform allows users to create personal blog websites on servers based on PHP and MySQL. A WordPress plugin is an application that can be added to a...

suricata-detections

suricata-detections Network IDS signature development grounde...

CVE-2026-42481

Open CASCADE Technology OCCT V800rc5 contains multiple vulnerabilities in its IGES and STEP file parsers that can be triggered by crafted IGES or STEP files. These issues include an out-of-bounds read in Geom2dBSplineCurve::EvalD0 during IGES B-spline curve evaluation, an out-of-bounds read in...

CVE-2026-41263 Traefik: BasicAuth middleware: timing side-channel vulnerability

Traefik is an HTTP reverse proxy and load balancer. Prior to versions 2.11.43, 3.6.14, and 3.7.0-rc.2, there is a timing side-channel vulnerability in Traefik's BasicAuth middleware that allows an attacker to enumerate valid usernames through response-time differences. The variable intended to ho...

Insertion Of Sensitive Information Into Sent Data

Axios is vulnerable to Insertion of Sensitive Information Into Sent Data. The vulnerability is due to improper use of truthy/falsy evaluation for the withXSRFToken configuration instead of strict boolean checks, which allows an attacker to force XSRF tokens to be sent to malicious cross-origin...

CVE-2026-42423

OpenClaw before 2026.4.8 contains an approval-timeout fallback mechanism that bypasses strictInlineEval explicit-approval requirements on gateway and node exec hosts. Attackers can exploit this timeout fallback to execute inline eval commands that should require explicit user approval,...

XekRung Technical Report

We present XekRung, a frontier large language model for cybersecurity, designed to provide comprehensive security capabilities. To achieve this, we develop diverse data synthesis pipelines tailored to the cybersecurity domain, enabling the scalable construction of high-quality training data and...

TwinGate: Stateful Defense against Decompositional Jailbreaks in Untraceable Traffic Via Asymmetric Contrastive Learning

Decompositional jailbreaks pose a critical threat to large language models LLMs by allowing adversaries to fragment a malicious objective into a sequence of individually benign queries that collectively reconstruct prohibited content. In real-world deployments, LLMs face a continuous, untraceable...

Jailbroken Frontier Models Retain Their Capabilities

As language model safeguards become more robust, attackers are pushed toward developing increasingly complex jailbreaks. Prior work has found that this complexity imposes a "jailbreak tax" that degrades the target model's task performance. We show that this tax scales inversely with model...

What to Look for in an Exposure Management Platform (And What Most of Them Get Wrong)

Every security team has a version of the same story. The quarter ends with hundreds of vulnerabilities closed. The dashboards are bursting with green. Then someone in a leadership meeting asks: "So, are we actually safer now?" Crickets. The room goes quiet because an honest answer requires contex...

ROS-20260429-73-0039

A vulnerability in the Intarray extension selectivity evaluation function of the PostgreSQL database management system is related to insufficient validation of the specified input data type. Exploitation of the vulnerability could allow an attacker acting remotely to execute arbitrary code in the...