Daily digest

China-Linked Hackers Backdoored Linux Login Software (PAM/OpenSSH) for Nearly a Decade

Sygnia researchers report that a China-nexus threat group tracked as Velvet Ant spent close to ten years hidden inside PAM and OpenSSH components on targeted Linux systems, giving them persistent authentication-layer access that survived routine remediation. The backdoor was planted at the login subsystem level, meaning standard endpoint cleanup would not remove it. The targeted network had no internet exposure, indicating deliberate, patient intrusion against hardened environments.

Why it matters: Nutanix AHV nodes and any Linux-based infrastructure (Ansible control planes, AWS GovCloud EC2 instances) rely on PAM and OpenSSH for privileged access — this TTP warrants auditing PAM stack integrity and SSH binary hashes on all Linux hosts, especially those handling CUI.

CISA Adds Oracle PeopleSoft Auth-Bypass CVE-2026-35273 to Known Exploited Vulnerabilities Catalog

CISA added CVE-2026-35273, a missing authentication vulnerability in Oracle PeopleSoft Enterprise PeopleTools, to its KEV catalog based on evidence of active exploitation. Federal agencies are subject to BOD 22-01 remediation deadlines for KEV entries. The flaw allows unauthenticated access to critical functions within PeopleSoft.

Why it matters: If PeopleSoft is in scope for your environment or any contractor you interface with, BOD 22-01 compliance requires tracking this KEV entry and confirming patch status — also relevant to CMMC practice CA.2.157 (identify and remediate vulnerabilities).

ShinyHunters Exploits Oracle Zero-Day to Steal Data from Higher Ed ERP Systems

The ShinyHunters threat group leveraged a zero-day vulnerability in Oracle's ERP software to exfiltrate large volumes of data from American universities. The bug disproportionately affected higher education institutions running Oracle ERP. Oracle has since patched the vulnerability, but data theft occurred before the fix.

Why it matters: Oracle products are common in DoD supply chains; if your org or any DIB partner runs Oracle ERP, verify patch status immediately — ShinyHunters is an active, capable group and this pattern of targeting enterprise software aligns with threats CMMC L2 environments face.

Over 400 Arch Linux AUR Packages Hijacked to Deploy Infostealer and eBPF Rootkit

Attackers compromised more than 400 packages in the Arch User Repository (AUR) by rewriting build scripts to install a Rust-based credential stealer on any machine that built the affected packages. When executed with root privileges, the malware can also load an eBPF rootkit to conceal itself. AUR is a community-maintained repository separate from Arch's official signed package mirrors.

Why it matters: If any developer workstations or build pipeline systems in your environment use Arch Linux or pull from AUR — including in containerized CI/CD workflows — credentials and access tokens on those systems should be treated as potentially compromised and rotated.

LangGraph Flaw Chain Exposes Self-Hosted AI Agents to Remote Code Execution

Researchers disclosed three now-patched vulnerabilities in LangGraph, an open-source LangChain framework for building multi-agent AI applications, including a critical SQL injection that chains into remote code execution. The flaws affect self-hosted deployments of LangGraph. Patches are available and were released alongside the disclosure.

Why it matters: If your self-hosted AI stack includes LangGraph-based agents, this is a critical patch — unauthenticated RCE on an AI agent host in a CMMC environment could constitute a reportable incident and a direct path to CUI exposure.

US Government Orders Anthropic to Block Foreign National Access to Fable 5 and Mythos 5 AI Models

The U.S. government issued an order directing Anthropic to suspend access to its most advanced AI models, Claude Fable 5 and Mythos 5, for all foreign nationals, citing national security concerns related to a jailbreak capability. Anthropic complied but publicly disputed the breadth of the order, arguing the cited jailbreak is narrow and the underlying capability is widely available. Both models were disabled worldwide pending compliance with the restriction.

Why it matters: Organizations using Anthropic models in GCC High or GovCloud workflows should assess whether service disruptions from government-mandated model suspensions are accounted for in continuity plans, and confirm that any AI tooling complies with emerging national security use restrictions.

CISA Revives Push to Finalize CIRCIA Cyber Incident Reporting Rules

CISA announced it will hold public meetings to advance the long-stalled Cyber Incident Reporting for Critical Infrastructure Act (CIRCIA) rulemaking, responding to pressure to finalize mandatory reporting requirements. The agency is also addressing widespread industry criticism of the draft regulations. CIRCIA will impose mandatory 72-hour incident reporting obligations on covered entities once finalized.

Why it matters: DIB contractors operating in CMMC L2 environments will likely fall under CIRCIA's covered entity definitions — now is the time to map your incident response procedures to the draft 72-hour reporting window before the rule finalizes.

Agentjacking Attack Tricks AI Coding Agents Into Executing Malicious Code via Fake Error Reports

Tenet Security researchers described a new attack class called "agentjacking" in which a manipulated Sentry error report causes AI coding agents to execute arbitrary code on a developer's machine. The attack exploits the trust AI agents place in tool outputs like error-tracking platforms. No CVE has been assigned as this is a design-level attack pattern rather than a single patchable flaw.

Microsoft Patches Surface Firmware Flaw That Allowed Devices to Be Bricked by a Single Packet

Microsoft has mostly resolved a firmware-level vulnerability in Surface hardware that allowed an unauthenticated attacker on the same network to send a single packet that would brick an unprotected device. The flaw was reportedly revealed inadvertently through Microsoft Copilot before a patch was ready. The fix is described as mostly complete, with further remediation still in progress.

Why it matters: If your Intune-managed fleet includes Surface devices, confirm firmware update deployment via Intune's driver/firmware update policies — a partially patched vulnerability in endpoint hardware on a CUI-handling device is a material risk under NIST 800-171 SI.3.218.

INTERPOL Operation Ramz Dismantles Sniper Dz Phishing-as-a-Service Platform, 201 Arrested

An INTERPOL-coordinated operation spanning October 2025 to February 2026 took down Sniper Dz, a decade-old phishing-as-a-service platform used to conduct large-scale credential theft campaigns. Authorities from 13 MENA-region countries participated, resulting in 201 arrests including the platform's primary administrator. Group-IB provided intelligence support for the operation.

CISA Infrastructure Security Chief Departs for ONCD Amid Ongoing Agency Restructuring

CISA's head of infrastructure security has moved to the Office of the National Cyber Director, adding to leadership turnover at the agency following a year of layoffs, buyouts, and restructuring. CISA is simultaneously preparing to hire hundreds of new employees. The leadership gap comes as the agency advances several major rulemaking and advisory initiatives.

Microsoft Resolves Windows Update Failures Caused by WUSA Installer When Deploying from Network Share

Microsoft has fixed a known issue causing Windows updates released since May 2025 to silently fail when deployed via the Windows Update Standalone Installer (WUSA) from a network share. The bug affected organizations using network-based update distribution rather than Windows Update or WSUS directly. A fix has been released and Microsoft recommends verifying update compliance on affected endpoints.

Why it matters: If your Windows 11 fleet uses Ansible or SCCM to deploy standalone update packages from a network share, over a year of patches may have silently failed to apply — audit patch compliance in Intune or your patch management tooling immediately to identify gaps that could affect CMMC SI.3.218 posture.