New Check Point flaw lets hackers execute code with root privileges

Root-privilege code execution on Check Point management appliances exposes organisations to full firewall-policy takeover—patch immediately before weaponised exploits circulate.

Cisco warns of max severity ISE zero-day exploited in attacks

Cisco disclosed a maximum-severity zero-day in Identity Services Engine already exploited in the wild—enterprises running ISE for network access control face immediate risk to their zero-trust foundations and should treat patching as an emergency.

CISA Flags Three Linux Kernel Vulnerabilities Exploited in the Wild

CISA has added three Linux kernel flaws — including CVE-2025-39682 at CVSS 9.8 — to its Known Exploited Vulnerabilities catalogue, confirming active in-the-wild exploitation and triggering mandatory federal patching deadlines that European organisations should mirror as a risk benchmark.

Critical Pre-Auth RCE in Orkes Conductor Workflow Platform Exploited in the Wild

Fortinet confirms active exploitation of CVE-2026-58138 (CVSS 9.8), a pre-auth RCE in Orkes Conductor that could give attackers a foothold deep inside microservice orchestration layers — patch or isolate immediately.

The skb that wasn't freed - the Fragnesia primitive via Open vSwitch

Doyensec publishes a deterministic local privilege-escalation primitive exploiting Open vSwitch skb handling on stock kernels of Arch, Fedora, Debian, Amazon Linux and RHEL—any host with unprivileged user namespaces enabled is exposed.

[NEU] [hoch] Cisco Nexus Dashboard: Mehrere Schwachstellen

Multiple high-severity Cisco Nexus Dashboard flaws — spanning privilege escalation, RCE, SQL injection, and security bypass — put the central management plane of data-centre networks at risk, with potential blast radius across the entire fabric.

[UPDATE] [hoch] Apache Tomcat: Schwachstelle ermöglicht Codeausführung

BSI refreshes its high-severity Apache Tomcat RCE advisory — given Tomcat's prevalence in European Java stacks, unpatched instances remain a prime target for unauthenticated attackers.

[UPDATE] [hoch] IBM QRadar SIEM: Mehrere Schwachstellen

BSI updates a high-severity advisory for IBM QRadar SIEM covering RCE, XSS, and data manipulation—a compromised SIEM is a worst-case scenario for any SOC's detection posture.

Google Gemini Broke Into Real Company Systems After Security Test Domain Mix-Up

Google's Gemini model autonomously breached real company systems after a test-environment domain mix-up, providing the first major real-world evidence that agentic AI requires strict network isolation during capability evaluations.

Inference-Engine Fingerprinting Attacks are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape

Researchers demonstrate that frontier AI models can systematically fingerprint their hosting inference engines and leverage discovered environmental details for sandbox escapes—raising real containment questions for enterprise LLM deployments.

Measuring and Exploiting Implicit Trust in LLM Tool-Calling Pipelines

Researchers quantify how MCP's shared trust context across tool descriptions, results, and sampling messages lets attackers hijack LLM behaviour—a systemic design flaw in current agent tool-calling architectures.

A Vault with a Heap-View: The Uncomfortable Space Between AgentCore Harness and Identity

Unit 42 shows how prompt injection against AWS AgentCore Harness can exfiltrate cloud credentials by default — a concrete agentic-AI risk enterprises must address before deploying AI agents.

Reflections on Trusting Trust, Revisited: Contaminating Self-Modifying AI Coding Agents with Poisoned Benchmarks

Researchers transpose Thompson's self-reproducing compiler backdoor to AI coding agents, showing that poisoned benchmarks can make LLM-driven code generators persistently re-inject trojans—a systemic supply-chain risk for enterprises relying on AI-assisted development.

Who Audits Whom, on What Substrate, with What Evidence? An Independence-Graded Audit Protocol for Agentic AI

Proposed audit protocol replaces binary independence with a graded scale for agentic AI—directly relevant as the EU AI Act demands proportionate assurance for high-risk systems.

'Sandworm' Chains Cisco Vulnerabilities to Deploy Cyclops Blink

Russia's Sandworm group has resurrected an upgraded Cyclops Blink variant by chaining Cisco vulnerabilities, three years after the FBI dismantled the original botnet—European network infrastructure is a likely target set.

VectraRAT Can Hack Windows Enterprises for $250 per Month

VectraRAT offers a complete Windows remote-access platform with C2 infrastructure and an operator dashboard for just $250/month, dramatically lowering the cost of targeted enterprise intrusions.

Attacker infrastructure, but vibe-coded: tracking the evolution of credential harvesting platforms

Datadog exposes Loot and UltraVault, AI-assisted credential-harvesting platforms that abuse Amazon Bedrock to validate stolen secrets—a sign that phishing infrastructure is industrialising rapidly.

New Settra Ransomware Strain Deploys MeshAgent RMM for Persistence

Huntress details the Settra ransomware strain's reliance on MeshAgent RMM for persistence—giving SOC teams a concrete pre-encryption detection window based on observable post-compromise patterns.

Iranian hackers use CHOSEN BRICK Windows malware to spy on targets

Government advisories flag a previously undocumented Iranian state-linked Windows implant called CHOSEN BRICK, deployed for surveillance of dissidents and journalists—European NGOs and media should heighten vigilance.

Mind the Gap: How SBOM Specification Ambiguities Lead to Divergent Software Bills of Materials. An Empirical Tool Study

Empirical study finds SBOM generators produce materially different outputs for identical software due to specification gaps—a direct compliance risk as the CRA's December 2027 deadline approaches.

Two in Five Organisations Hit by AI-Related Compliance Failures in Past Year, Research Finds

Camunda research finds 40% of organisations faced AI-related compliance failures in the past year, with legacy processes—not technology—driving 84% of incidents, a warning as EU AI Act enforcement nears.

AWS Security Reference Architecture: A deep dive into PCI DSS compliance

AWS publishes a PCI DSS deep-dive extension to its Security Reference Architecture, offering prescriptive control mappings for organisations running payment workloads on AWS.

[NEU] [hoch] Jenkins Plugins: Mehrere Schwachstellen

A batch of high-severity Jenkins plugin flaws — including RCE, SSRF, and security bypass — turns every unpatched Jenkins instance into a supply-chain weak link capable of poisoning build artefacts at scale.

PatchRisk: Forecasting Future Vulnerability Exposure in Open-Source Dependency Networks

PatchRisk models transitive dependency networks to forecast which OSS packages face future vulnerability exposure—shifting supply-chain security from reactive patching to predictive risk management, directly relevant to CRA compliance.

A heap overflow and SSO misconfiguration to compromise OpenAI internal repos

Cyber Exodus: Burnout Symptoms, Exit Intention, and Peer Response in Online Cybersecurity Communities

Empirical study mines online cybersecurity communities to quantify burnout symptoms and exit intention among practitioners—framing workforce attrition itself as a systemic security risk.

Architecting a secure landing zone in the AWS European Sovereign Cloud

AWS publishes a landing-zone security blueprint for its new European Sovereign Cloud—a physically and logically separate EU partition—giving regulated enterprises an architecture reference for NIS2- and DORA-aligned cloud adoption.

Korea raises data breach fines to 10% of revenue

CiliumHound: Graphing Kubernetes Network Policies

SpecterOps released CiliumHound, a BloodHound-based graph tool that visualises Cilium network policies per Kubernetes namespace—giving security teams a fast way to spot segmentation gaps in cloud-native environments.