Endpoint Security

GhostWrite Vulnerability Facilitates Attacks on Devices With RISC-V CPU

Researchers disclose the details of GhostWrite, a RISC-V CPU vulnerability that can be exploited to gain full access to targeted devices.

GhostWrite

LAS VEGAS — BLACK HAT USA 2024 — A team of researchers from the CISPA Helmholtz Center for Information Security in Germany has disclosed the details of a new vulnerability affecting a popular CPU that is based on the RISC-V architecture. 

RISC-V is an open source instruction set architecture (ISA) designed for developing custom processors for various types of applications, including embedded systems, microcontrollers, data centers, and high-performance computers. 

The CISPA researchers have discovered a vulnerability in the XuanTie C910 CPU made by Chinese chip company T-Head. According to the experts, the XuanTie C910 is one of the fastest RISC-V CPUs.

The flaw, dubbed GhostWrite, allows attackers with limited privileges to read and write from and to physical memory, potentially enabling them to gain full and unrestricted access to the targeted device.

While the GhostWrite vulnerability is specific to the XuanTie C910 CPU, several types of systems have been confirmed to be impacted, including PCs, laptops, containers, and VMs in cloud servers.  

The list of vulnerable devices named by the researchers includes Scaleway Elastic Metal RV bare-metal cloud instances; Sipeed Lichee Pi 4A, Milk-V Meles and BeagleV-Ahead single-board computers (SBCs); as well as some Lichee compute clusters, laptops, and gaming consoles. 

Advertisement. Scroll to continue reading.

“To exploit the vulnerability an attacker needs to execute unprivileged code on the vulnerable CPU. This is a threat on multi-user and cloud systems or when untrusted code is executed, even in containers or virtual machines,” the researchers explained. 

To demonstrate their findings, the researchers showed how an attacker could exploit GhostWrite to gain root privileges or to obtain an administrator password from memory.

Unlike many of the previously disclosed CPU attacks, GhostWrite is not a side-channel nor a transient execution attack, but an architectural bug.

The researchers reported their findings to T-Head, but it’s unclear if any action is being taken by the vendor. SecurityWeek reached out to T-Head’s parent company Alibaba for comment days before this article was published, but it has not heard back. 

Cloud computing and web hosting company Scaleway has also been notified and the researchers say the company is providing mitigations to customers. 

It’s worth noting that the vulnerability is a hardware bug that cannot be fixed with software updates or patches. Disabling the vector extension in the CPU mitigates attacks, but also impacts performance.

The researchers told SecurityWeek that a CVE identifier has yet to be assigned to the GhostWrite vulnerability. 

While there is no indication that the vulnerability has been exploited in the wild, the CISPA researchers noted that currently there are no specific tools or methods for detecting attacks. 

Additional technical information is available in the paper published by the researchers. They are also releasing an open source framework named RISCVuzz that was used to discover GhostWrite and other RISC-V CPU vulnerabilities. 

Related: Intel Says No New Mitigations Required for Indirector CPU Attack

Related: New TikTag Attack Targets Arm CPU Security Feature 

Related: Researchers Resurrect Spectre v2 Attack Against Intel CPUs

Related Content

Endpoint Security

Branch Target Reuse (BTR) is a new Spectre v2 attack targeting JIT compilers in web browsers, language runtimes, and the operating system kernel

Endpoint Security

Researchers have disclosed technical details on a new AMD processor attack that allows remote code execution inside confidential VMs.

Endpoint Security

The PCIe flaws, found by Intel employees, can be exploited for information disclosure, escalation of privilege, or DoS.

Cloud Security

Intel and AMD have published advisories after academics disclosed details of the new TEE.fail attack method.

Cloud Security

A vulnerability in RMP initialization allows the AMD processor’s x86 cores to maliciously control parts of the initial RMP state.

Endpoint Security

The attack uses a passive interposer to control the SGX enclave and extract the DCAP attestation key, breaking the mechanism.

Cloud Security

L1TF Reloaded is a vulnerability combining the old L1TF and half-Spectre hardware flaws to bypass deployed software mitigations.

Cloud Security

Exploiting incomplete speculative execution attack mitigations extended to the branch predictor state, VMScape leaks arbitrary memory.

Copyright © 2026 SecurityWeek ®, a Wired Business Media Publication. All Rights Reserved.

Exit mobile version