Follow us on Google News
Google Cloud has warned that three security vulnerabilities affecting AMD’s fifth-generation EPYC processors, known as Turin, could allow attackers to infer sensitive information from other processes or the operating system kernel within virtual machines.
The vulnerabilities affect several Google Cloud machine series and involve processor-level weaknesses known as Transient Scheduler Attacks (TSA). Although Google has already implemented protections against attacks between virtual machines and their hosts, additional security updates are required to address potential information leakage within individual virtual machines.
Google has classified the issue as medium severity and advised affected customers to install operating system patches when available. Customers must also stop and restart their virtual machines after receiving the necessary CPU information updates.
FCRF’s Chief AI Officer Certification Returns on November 14 to Prepare Future AI Leaders
How Could the AMD Processor Vulnerabilities Expose Sensitive Data?
The three vulnerabilities are identified as CVE-2024-36350, CVE-2024-36357 and CVE-2026-0004.
The first two, known as TSA-SQ and TSA-L1, were initially disclosed in July 2025 in connection with earlier generations of AMD EPYC processors.
However, AMD has now confirmed that its Zen 5 processors, including the Turin family, are affected differently from previously examined processors.
The newly disclosed vulnerability, CVE-2026-0004, involves potential information leakage from the processor’s Level 1 data cache, which temporarily stores information for rapid access by the CPU.
Under certain conditions, an attacker capable of executing code on an affected system could exploit these weaknesses to infer data associated with other processes running on the same processor core.
The Xen Project, which maintains the Xen hypervisor, has identified the new vulnerability as TSA-Fill.
AMD has classified the vulnerabilities as local attacks involving high exploitation complexity and requiring low privileges.
Google has not reported evidence of active exploitation. None of the three vulnerabilities was listed in the US Cybersecurity and Infrastructure Security Agency’s Known Exploited Vulnerabilities catalogue at the time of the advisory.
Which Google Cloud Virtual Machines Are Affected?
According to Google’s security bulletin, the affected machine series include C4D, N4D, H4D, Z4D and G4.
The advisory also covers bare-metal instances of the C4D and Z4D series.
Google said it has already applied mitigations intended to prevent attacks between separate virtual machines and between virtual machines and their underlying hosts.
However, attacks involving processes operating within the same virtual machine require additional protections at the guest operating system level.
To enable these protections, Google is updating the CPU information presented to affected virtual machines so that patched operating system kernels can activate the relevant security mitigations.
Customers must install the appropriate operating system fixes once released by their vendors.
After Google’s CPU information update reaches an affected virtual machine, administrators must stop and start the instance to activate the necessary changes.
Google specifically warned that an ordinary guest operating system reboot would not be sufficient.
For customers running their own operating systems or hypervisors on affected bare-metal instances, the company recommends applying the relevant vendor security updates.
Google also plans to provide updated Google Kubernetes Engine node images for affected Turin-based bare-metal systems. Customers using Cloud Run do not need to take action under this advisory.
What Security Updates Are Available for AMD Turin Systems?
Security fixes addressing the vulnerabilities are being introduced through operating system kernels, platform firmware and hypervisor updates.
A Linux kernel change designed to mitigate Transient Scheduler Attacks affecting Zen 5 processors, including the newly identified variant, has already been incorporated into the Linux kernel source tree.
However, the availability of fixes varies across operating system distributions.
At the time of the advisory, Debian had addressed CVE-2026-0004 in the Linux kernel package for its unstable branch, while the security tracker continued to list affected kernels in the bookworm, trixie and forky branches as vulnerable.
The Xen Project has also released hypervisor patches addressing the security weaknesses. It stated that microcode covering the TSA vulnerabilities on Zen 5 processors had been included in linux-firmware since September 23.
For organisations operating AMD Turin processors on their own physical servers, AMD recommends both platform firmware and operating system updates to provide complete mitigation.
The company has cautioned that enabling security protections may affect system performance and advised administrators to assess the implications for their workloads.
Google has not specified how customers can determine when the updated CPU information has reached their virtual machines.
The advisory highlights the importance of applying processor-level security fixes alongside operating system and infrastructure updates, particularly in cloud environments where several applications and processes may share underlying computing resources.
About the author — Suvedita Nath is a science student with a growing interest in cybercrime and digital safety. She writes on online activity, cyber threats, and technology-driven risks. Her work focuses on clarity, accuracy, and public awareness.
