Daily digest

16-Year-Old Linux KVM Flaw Lets Guest VMs Escape to Host on Intel and AMD x86 Systems

CVE-2026-53359, dubbed 'Januscape,' is a use-after-free bug in the Linux KVM shadow MMU code affecting both Intel and AMD x86 systems. A guest VM can trigger the flaw to corrupt host kernel state; a public proof-of-concept reliably panics the host, and the researcher claims a separate unreleased exploit achieves arbitrary code execution on the host. The vulnerability has existed in the kernel for approximately 16 years.

Why it matters: Nutanix AHV is built on KVM — a guest-to-host escape means a compromised VM could reach the hypervisor and every other tenant VM on the node. Patch the underlying Linux kernel on all AHV hosts immediately and review Nutanix's advisory for a specific fix or workaround.

CitrixBleed-ing Again? NetScaler Vulnerability Under Attack

Attackers are actively exploiting a new memory disclosure vulnerability in Citrix NetScaler products shortly after a public proof-of-concept was released. The flaw echoes the 2023 CitrixBleed incident in class and exploitation speed. Vulnerable NetScaler ADC and Gateway instances are being targeted in the wild.

Why it matters: If NetScaler appliances sit anywhere in your network perimeter or VPN stack, patch immediately — the PoC lowers the bar for exploitation and the original CitrixBleed led to widespread session-token theft against government and defense contractors.

BeyondTrust Patches Critical Auth Bypass Flaws in Remote Support and PRA

BeyondTrust released patches for two critical pre-authentication vulnerabilities in its Remote Support (RS) and Privileged Remote Access (PRA) products, tracked as CVE-2026-40138 (CVSS 9.2) and a second related flaw. Unauthenticated attackers can exploit these to take control of affected systems. Both products are widely used for privileged session management in enterprise environments.

Why it matters: BeyondTrust PRA is commonly used in CMMC-scope environments for privileged access management (NIST 800-171 AC.2.006/AC.3.017). An unauthenticated bypass of your PAM tool is a critical-path compromise — apply patches and review session logs for anomalous access immediately.

Fake IT Support Calls on Microsoft Teams Push EtherRAT Malware

Unit 42 documented a campaign where threat actors pose as corporate IT helpdesk staff over Microsoft Teams voice calls, socially engineering employees into granting remote control of their machines. Once access is granted, attackers drop the EtherRAT remote access trojan to establish persistent footholds in corporate networks. The technique bypasses many email-based phishing controls by abusing a trusted collaboration platform.

Why it matters: Your M365 GCC High tenant's Teams environment is a direct attack surface for this technique. Review external access and guest federation settings in Teams Admin Center, and reinforce user training that the real helpdesk will never call via Teams and request remote control.

Threat Actors Probe Gitea Docker Flaw CVE-2026-20896 13 Days After Disclosure

Active exploitation attempts against CVE-2026-20896 (CVSS 9.8) in Gitea Docker images were observed by Sysdig only 13 days after the vulnerability was publicly disclosed. The flaw allows unauthenticated remote attackers to gain elevated privileges by sending a crafted 'X-WEBAUTH-USER' HTTP header, which Gitea trusts from any source IP. Self-hosted Gitea Docker deployments are the primary target.

Why it matters: If you run a self-hosted Gitea instance in containers for source control or CI/CD pipelines, this is under active exploitation — update the Docker image immediately and verify the instance is not exposed to untrusted networks.

Max Severity Adobe ColdFusion Flaw Now Exploited in Attacks

CVE-2026-48282, a maximum-severity vulnerability in Adobe ColdFusion, is now being actively exploited in the wild according to vulnerability intelligence firm KEVIntel. The flaw was previously disclosed by Adobe with a patch available, but active exploitation signals threat actors are prioritizing it against unpatched systems. ColdFusion is commonly deployed in legacy web application stacks across government and enterprise environments.

CERT/CC Warns of Hidden Admin Backdoor in Tenda Router Firmware

CERT/CC issued an advisory for CVE-2026-11405, an undocumented hardcoded authentication backdoor embedded in multiple Tenda router firmware versions that grants full administrative access to the web management interface without valid credentials. The backdoor allows any attacker who can reach the management interface to bypass password verification entirely. No patch from Tenda has been announced.

Why it matters: Tenda devices occasionally appear in lab, out-of-band management, or small-office segments. Audit your network inventory for any Tenda routers; with no patch available, immediate isolation or replacement is the only mitigation.

Microsoft to Enable Windows Settings Backup by Default for Orgs

Microsoft announced that the Windows settings backup and restore feature will be enabled by default on Microsoft Entra-joined and Entra hybrid-joined devices after upgrading to Windows 11 26H2. Settings data will sync to users' OneDrive accounts as part of the backup process. Administrators can disable or configure the feature via policy.

Why it matters: In a GCC High / CMMC environment, automatic syncing of settings data to OneDrive raises data boundary questions — confirm that backup data flows to your GCC High OneDrive tenant and not commercial, and evaluate whether the feature should be explicitly scoped or disabled via Intune configuration policy before 26H2 rollout.

CISA Expects to Finalize Key Cyber Reporting Rule by September

CISA confirmed it expects to publish the final rule implementing CIRCIA (Cyber Incident Reporting for Critical Infrastructure Act) by September 2026, following additional stakeholder town halls held in June after a DHS funding lapse delayed the process earlier this year. The final rule will mandate cyber incident reporting timelines and scope for covered entities across critical infrastructure sectors. The rulemaking has been in progress since CIRCIA was signed in 2022.

Why it matters: Defense contractors and government IT operators should finalize internal incident response and reporting procedures now — CIRCIA's final rule will impose specific reporting windows that must be operationalized before the rule takes effect, and gaps could compound CMMC compliance obligations.

FedRAMP 2026 Is Not a Compliance Update — It's a New Operating Model

A former DHS CISO argues in FedScoop that FedRAMP's 2026 changes represent a structural shift from point-in-time, documentation-based compliance toward continuous evidence collection and real-time authorization status. Cloud service providers and agencies will need to maintain live, auditable control evidence rather than periodic assessment artifacts. The piece outlines the operational and tooling implications for both CSPs and government IT consumers.

Why it matters: If your AWS GovCloud workloads or M365 GCC High services rely on FedRAMP authorizations as a compliance backstop, understand that the authorization posture of those services will increasingly be dynamic — your own continuous monitoring program (required under NIST 800-171 CA.2.157) should align accordingly.

France to Stop Certifying Non-Quantum-Safe Encryption

France's national cybersecurity agency ANSSI announced it will halt certification of security products that do not incorporate quantum-resistant encryption algorithms. The policy will compel government bodies and critical infrastructure operators in France to migrate away from classical cryptographic systems. ANSSI framed the move as a proactive step given projected timelines for cryptographically relevant quantum computers.

Enforce Least-Privilege Authorization in Multi-Agent AI Chains Using Cedar

AWS published a technical guide on using Cedar, its open-source policy language, to enforce least-privilege authorization across multi-agent AI workflows where agents delegate tasks through multi-hop chains. The post addresses an OWASP Top 10 risk for agentic applications: authorization scope silently expanding beyond what the originating user intended as agents pass tasks downstream. Cedar policies can constrain what each agent in a chain is permitted to do regardless of what the upstream agent requests.

Why it matters: If you are deploying or planning agentic AI workloads on AWS GovCloud, this pattern directly addresses a NIST 800-171 AC least-privilege control gap that is difficult to satisfy with traditional RBAC alone in multi-step AI pipelines.