Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations

In this paper, we investigate whether refusal behavior can be predicted from LLM intermediate activations before decoding using linear probes trained on residual stream activations at each transformer block. We find that refusal is linearly decodable well before the final layer, indicating that...

The Fault in Our Drafts: Vulnerabilities in RPKI Specification and Software

The Resource Public Key Infrastructure RPKI secures the Internet's routing system by defining a complex trust and validation framework for certificates, Route Origin Authorizations ROAs, manifests, and Certificate Revocation Lists CRLs. These mechanisms are specified across dozens of RFCs. This...

Ghost CMS CVE-2026-26980 Exploited to Hijack 700+ Sites for ClickFix Attacks

Threat actors are exploiting a recently disclosed critical security flaw in Ghost CMS to inject malicious JavaScript code with an aim to fuel ClickFix attacks. According to QiAnXin XLab, the activity involves the exploitation of CVE-2026-26980 CVSS score: 9.4, an SQL injection vulnerability in...

A Hacker Group Is Poisoning Open Source Code at an Unprecedented Scale

GitHub is just the latest victim of TeamPCP, a gang that has carried out a spree of software supply chain attacks that has impacted hundreds of organizations...

EUVD-2026-31133

InfoScale CmdServer before 7.4.2 mishandles access control...

CVE-2026-44924

InfoScale VIOM 9.1.3 allows XSS...

Astra Linux - уязвимость в linux-5.10, linux-6.1, linux-5.15

In the Linux kernel, the following vulnerability has been resolved: rcuscale: Move rcuscalewriter’s scheduletimeoutuninterruptible function to idle. The rcuscale.holdoff module parameter can be used to delay the start of rcuscalewriter’s kthread. However, the hung-task timeout will trigger when t...

Malicious code in @antv/word-scale-chart (npm)

Part of the Mini Shai-Hulud supply chain attack campaign in which a threat actor compromised the npm account atool and published 631 malicious versions across 314 npm packages in an automated 22-minute burst. Each malicious version injects a preinstall hook that executes a 498KB obfuscated Bun...

MAL-2026-4083 Malicious code in @antv/scale (npm)

Part of the Mini Shai-Hulud supply chain attack campaign in which a threat actor compromised the npm account atool and published 631 malicious versions across 314 npm packages in an automated 22-minute burst. Each malicious version injects a preinstall hook that executes a 498KB obfuscated Bun...

1byte-react-design (>=1.7.1 <=1.14.0), 1g6table (=0.1.0) +2674 more potentially affected by unknown CVE via @antv/scale (>=0.0.1 <=0.5.2)

@antv/scale NPM version =0.0.1, =1.7.1, =1.1.0, =0.1.1, =0.1.1, =0.1.0, =0.0.2, =0.0.1-beta, =0.1.2, =1.1.43, =5.0.48, =1.0.1, =1.1.44 and more Source cves: unknown CVE Source advisory: OSV:MAL-2026-4083...

@antv/narrative-text-vis (>=0.1.6 <=0.2.5), antv-site-demo-rc (>=0.1.0-alpha.16 <=0.1.0-alpha.22) potentially affected by unknown CVE via @antv/word-scale-chart (>=0.1.1 <=0.3.4)

@antv/word-scale-chart NPM version =0.1.1, =0.1.6, =0.1.0-alpha.16, =0.1.0-alpha.22 Source cves: unknown CVE Source advisory: OSV:MAL-2026-4096...

MemRepair: Hierarchical Memory for Agentic Repository-Level Vulnerability Repair

Modern software ecosystems face a rapidly growing number of disclosed vulnerabilities, increasing the need for automated repair techniques that can operate reliably at repository scale. Although Large Language Model LLM-based agents have recently shown promise for automated vulnerability repair...

ADR: An Agentic Detection System for Enterprise Agentic AI Security

We present the Agentic AI Detection and Response ADR system, the first large-scale, production-proven enterprise framework for securing AI agents operating through the Model Context Protocol MCP. We identify three persistent challenges in this domain: 1 limited observability -- existing Endpoint...

The End of Trust: How Agentic AI Breaks Security Assumptions

For decades, the security of digital interaction has rested on an unacknowledged economic constraint. Attackers faced a tradeoff between the fidelity of a deception and the scale at which it could be deployed. Convincing impersonation required sustained human effort and was confined to a narrow s...

Defense at AI speed: Microsoft’s new multi-model agentic security system tops leading industry benchmark

In this article 1. AI-powered vulnerability discovery at hyper-scale 2. Codename: MDASH—Microsoft Security’s new multi-model agentic scanning harness 3. Using codename MDASH for security research 4. The 5.12.2026 Patch Tuesday cohort 5. Two deep dives 1. CVE-2026-33827—Remote unauthenticated UAF ...

Security Bulletin: IBM Storage Scale System: Vulnerability in Linux kernel crypto subsystem could allow local privilege escalation (CVE-2026-31431)

Summary IBM Storage Scale Systems is affected by a security vulnerability identified in the Linux kernel's cryptographic interface CVE-2026-31431 that could allow a local user with low privileges to escalate to root privileges. The vulnerability has a CVSS score of 7.8 High and requires local...

CVE-2026-26946

Dell ECS versions 3.8.1.0 through 3.8.1.7 and Dell ObjectScale versions prior to 4.3.0.0, contains an improper privilege management vulnerability in the OS. A high privileged attacker with local access could potentially exploit this vulnerability, leading to elevation of privileges...

CVE-2025-43992

Dell ECS versions 3.8.1.0 through 3.8.1.7 and Dell ObjectScale versions prior to 4.3.0.0, contains an authentication bypass by assumed-immutable data vulnerability in Geo replication. An unauthenticated attacker with remote access could potentially exploit this vulnerability, leading to...

CVE-2026-40636

Dell ECS versions 3.8.1.0 through 3.8.1.7 and Dell ObjectScale versions prior to 4.3.0.0, contains a use of hard-coded credentials vulnerability. An unauthenticated attacker with local access could potentially exploit this vulnerability, leading to filesystem access for attacker...

Dell ECS 安全漏洞

Dell ECS is an enterprise-level object storage solution from the American company Dell. Versions 3.8.1.0 to 3.8.1.7 of Dell ECS, as well as versions prior to 4.3.0.0 of Dell ObjectScale, have security vulnerabilities. These vulnerabilities stem from a certification bypass in Geo replication, whic...