Hello, stay ahead with CISO Brief 🚀

🚀 Amazon Aurora and Amazon RDS now support AWS Graviton5-based R9g database instances for Aurora with PostgreSQL and MySQL compatibility and for RDS PostgreSQL, MySQL, and MariaDB. R9g instances deliver up to a 30% performance improvement and up to a 22% price-performance gain over Graviton4 equivalents, with configurations ranging from large to 48xlarge. Built on the AWS Nitro System, they provide up to 192 vCPUs, an 8:1 memory-to-vCPU ratio, high network bandwidth, and enhanced EBS throughput, and are available in select US and EU regions.

📰 This edition of the Threat Source newsletter reflects on the value of asking questions to foster understanding and collaboration across cultures and teams. It highlights Cisco Talos' analysis of a massive Microsoft Patch Tuesday (November 2026) addressing 1,095 vulnerabilities, including 64 critical flaws and an actively exploited Outlook spoofing bug (CVE-2026-124194). The update also calls attention to a Google Chrome zero-day (CVE-2026-125350) and urges a targeted, risk-based patching approach, while noting Snort rules to help mitigate immediate threats.

🔒 Chrome Enterprise is extending protections to clinical endpoints as healthcare adopts web-based apps and AI agents. The offering is now available in the IGEL Marketplace and supports Citrix UniconOS, enabling secure native browsing across diverse devices. Collaborations with Omnissa add device trust and zero-trust posture checks, while new features optimize Epic Hyperspace workflows with silent printing and native PDF handling. These updates aim to help IT teams deploy modern clinical technology while maintaining DLP and governance.

🛡️ This ThreatsDay bulletin surveys a week of cyber incidents where attackers used fake sites, stolen access, trusted tools and AI to bypass defenses. Highlights include a sophisticated North Korean crypto-laundering operation, a surge in GSocket detections on Linux servers, and a Europol takedown seizing 235 malicious servers. Also notable are browser-based espionage via the Falcon JavaScript C2, discovery of 59 Linux kernel issues, and new AI-enabled and OAuth-based attack techniques.

🔐 Amazon Managed Workflows for Apache Airflow (Amazon MWAA) now offers fine-grained access control through IAM Identity Center integration and multi-team environments. Users sign in with their corporate identity provider and authorization is evaluated per request using IAM policies and condition keys that scope access by resource type, action, and resource ID. Multi-team environments let multiple teams share one MWAA environment with separate S3 DAG folders and visibility based on IAM Identity Center group membership.

🛡️ A public proof-of-concept called SignDrop claims a local Windows user can gain SYSTEM privileges by abusing a log authentication feature in the Common Log File System driver (clfs.sys). The exploit uses object manager symbolic links to redirect creation of companion .cnpf files so they are created in a privileged process's security context, enabling a writable SYSTEM-owned file. Microsoft has not yet confirmed the flaw or issued a CVE or patch.

🔒 Google Cloud and SentinelOne announced a re-architected integration to deliver endpoint telemetry to Google Security Operations with lower latency and reduced operational cost. The collaboration replaces intermediate storage hops with direct API streaming, reduces event sizes via a modernized EDR schema, and provides precision telemetry scoping through SentinelOne Cloud Funnel. Engineering efforts also align alert models and SOAR playbooks with Mandiant teams for automated response and enriched investigations.

🔒 Organization Policy Service helps enforce resource guardrails across Google Cloud using a catalog of predefined policies and Custom Organization Policy. The blog announces GA of the Organization Policy Model Context Protocol (MCP) server and natural-language constraint authoring powered by Gemini, enabling agents and pipelines to interact with policies directly. Features include constraint discovery, policy evaluation, and management tools like CreatePolicy and CreateCustomConstraint, plus testing support for compliant and non-compliant configurations.

🔗 Google Cloud is expanding its Cross-Cloud Interconnect to include last-mile connectivity, beginning a preview with Lumen Technologies. The service uses open interconnect APIs to automate provisioning and deliver dedicated Layer 3 private connectivity to Google Cloud in minutes rather than months. This aims to simplify connections to private clouds, neoclouds, on-premises data centers, and other clouds, using Lumen’s last-mile network and a single operational runbook for multicloud links.

🛡️ Two Microsoft teams demonstrate how post-training supply-chain attacks can implant latent backdoors in coding and terminal agents. Dataset poisoning hides crafted examples in fine-tuning data to teach agents to identify and misuse available command-execution tools, achieving ~78% success in-distribution. Model poisoning assumes control of training and distributed checkpoints, combining SFT and RL to reach 84–95% success on unseen repositories. Both attacks exploit trust in datasets and model weights and highlight the need for data scrutiny, checkpoint provenance, runtime evaluation, and strict tool permissions.

🔧 Amazon Connect Customer now enables agents to merge and link related cases directly within the agent workspace to streamline issue resolution and maintain consistent responses. Agents can consolidate histories when a customer opens multiple cases for the same issue, or link cases across different customers affected by the same problem to coordinate responses. The feature is available in multiple AWS regions and integrates into the existing Cases functionality.

🛡️Meta’s new AI agent Muse offers wide-ranging assistant features — from message drafting to making calls — but early adopters and researchers rapidly discovered privacy and security problems. A journalist found Muse accessing local conversations without clear permission, while a researcher disclosed a vulnerability that let attackers redirect voice processing to malicious servers and seize control of the agent. Meta also tested routing Muse’s automated calls to human operators without clearly informing employees, raising further privacy concerns. Meta has released fixes and guidance, and users should review permissions and data-training settings to limit exposure.

🔒 A critical flaw in Red Hat Quay lets an attacker with access to a Quay registry read and overwrite private container images across repositories. Anonymous access is enabled by default, allowing unauthenticated exploitation; when disabled, any valid account can still exploit the issue. Red Hat released fixed builds for eight Quay 3 branches and tracked the key issue as CVE-2026-123710 (CVSS 9.8). A second critical bug, CVE-2026-110697 (CVSS 9.9), enables tenant token theft via namespace wildcard handling in robot-account listing.

🚀 Google Cloud is recognized by Gartner as a Leader and placed furthest in vision in the inaugural Magic Quadrant for Data Lakehouse Platforms. The post argues that the enterprise lakehouse has evolved into a foundation for autonomous AI agents, and presents Google’s borderless Lakehouse — built on Apache Iceberg — as a unified, governed fabric that spans multi-cloud, SaaS, and operational data. It highlights integrations with BigQuery, Gemini, and Knowledge Catalog, customer examples like Commerzbank and PayPal, and cost controls that reduce token consumption for large-scale agentic workloads.

📣 Amazon Simple Notification Service (Amazon SNS) now supports sending push notifications from four additional AWS Regions: Asia Pacific (Hyderabad), Asia Pacific (Melbourne), Europe (Zurich), and AWS GovCloud (US-East). The fully managed pub/sub service delivers messages to endpoints including AWS Lambda, Amazon SQS, HTTP, SMS, email, and push services such as APNs, FCM, ADM, Baidu, and WNS. Customers in these Regions can now send push notifications using the same supported providers as other Regions.

🔒 Citrix has released patches addressing multiple critical vulnerabilities in NetScaler ADC, NetScaler Gateway, NetScaler Console, and Secure Access Client for Windows and Linux. The fixes cover several high-severity memory overflow and corruption bugs that could cause denial-of-service or, in one case, authentication bypass, as well as privilege escalation issues in client software. Customers are urged to apply the listed updates to mitigate these risks.

🔍 Cybersecurity researchers from Island uncovered a campaign dubbed TypoPrompting where 186 lookalike domains impersonating 27 AI brands redirect users to ClickFix-style lures, browser-hijacking malware, and search-hijack extensions. The fake sites use fingerprinting and fake verification pages to run PowerShell or shell commands, deliver the Net Guard hijacker, or push malicious Chrome extensions with nearly one million users. Several domains were registered as early as July 2023 and the infrastructure leverages parked-domain monetization and traffic distribution systems to weaponize legitimate ad flows.

🔍 Unit 42 at Palo Alto Networks reports a large China-based campaign that used the open-source AI penetration framework Cairn with Chinese models Zhipu GLM and DeepSeek to autonomously run 25 operations against at least 19 organizations in Aug–Sep 2026. The attackers achieved unauthorized access at 10 businesses, exfiltrated credentials and records, created persistent admin accounts, and altered critical artifacts including a live industry ballot. Unit 42 detailed the attack chain, the shared-blackboard coordination model and how attackers automated authorization pretexts to evade defenses.