Critical UEFI DMA Vulnerability Lets PCIe Devices Bypass Boot Security on Popular Motherboards

CyberSecureFox 🦊

Researchers at Riot Games, in coordination with CERT/CC, have identified a critical UEFI firmware vulnerability in several motherboard lines from Asus, Gigabyte, MSI and ASRock. The flaw allows a malicious PCIe device to perform a Direct Memory Access (DMA) attack during early boot, bypassing key security mechanisms before the operating system is loaded. Due to differences between vendors’ implementations, the issue has been assigned multiple identifiers: CVE-2025-11901, CVE-2025-14302, CVE-2025-14303 and CVE-2025-14304.

What Is DMA and Why IOMMU Is Critical for Memory Security

Direct Memory Access (DMA) is a hardware feature that lets devices such as GPUs, Thunderbolt controllers and PCIe expansion cards read from and write to system RAM without going through the CPU for every transaction. This design significantly improves performance for I/O-heavy workloads, but it also creates a powerful primitive for attackers if access is not tightly controlled.

The component responsible for controlling this access is the IOMMU (Input-Output Memory Management Unit). In simple terms, it acts as a firewall for RAM, defining which regions of memory each device is allowed to interact with and blocking unauthorized access. When configured correctly, the IOMMU enforces DMA remapping, limiting what a compromised or malicious device can see or modify in system memory.

How the UEFI DMA Vulnerability Works

According to Riot Games and CERT/CC, the core problem lies in how certain UEFI implementations initialize the IOMMU during the earliest stages of the boot process. At this point, the firmware is responsible for enabling and properly configuring the IOMMU before any device is allowed to execute DMA operations. This small window in the boot sequence is crucial for establishing a trustworthy environment.

In affected firmware, UEFI reports that DMA protection is enabled, but in practice the IOMMU is not correctly initialized when control is passed to the next stage in the boot chain. As a result, any physically attached DMA-capable PCIe device can read and modify system memory before the operating system and its security controls are active. This is comparable in spirit to earlier academic work on Thunderbolt-based DMA attacks (such as the Thunderclap research), but triggered by firmware misconfiguration rather than interface design alone.

Why Pre-Boot DMA Attacks Are Especially Dangerous

Attacks that execute before the OS loads are among the most dangerous classes of compromise. If an adversary injects malicious code into memory at this stage, they can:

  • tamper with the boot chain, UEFI drivers and early OS components;
  • install stealthy bootkits that persist across reboots;
  • hide from endpoint security tools that assume a trustworthy boot environment;
  • bypass OS-level protections that depend on accurate IOMMU configuration.

These capabilities make pre-boot DMA attacks attractive to advanced threat actors targeting high-value systems, not just opportunistic attackers.

Riot Games’ Response: Vanguard Blocks Valorant on Vulnerable Systems

BleepingComputer reports that Riot Games was the first to encounter practical consequences of this vulnerability in the wild. The company’s kernel-level anti-cheat solution, Vanguard, began refusing to start on systems where DMA protections during boot could not be trusted. From an anti-cheat perspective, if a cheat is able to load before Vanguard, it can attempt to hide within firmware or low-level drivers that Vanguard cannot reliably inspect.

Riot has updated Vanguard so that, when the platform appears vulnerable to early DMA attacks, Valorant simply does not launch. Instead, users see a notification explaining the risk, with guidance to update their motherboard’s UEFI/BIOS firmware and review security-related settings such as IOMMU and virtualization options.

Affected Hardware, Threat Model and Real-World Risk

Physical Access as the Primary Exploitation Requirement

CERT/CC confirms the flaw for selected boards from ASRock, Asus, Gigabyte and MSI, and similar issues may exist in other vendors’ firmware as well. Importantly, exploiting this class of bug requires physical access to the target machine: the attacker must connect a malicious PCIe or other DMA-capable device that can launch the attack during boot.

This aligns with scenarios involving “attacker with physical access” — for example, shared or public workstations, coworking spaces, hotel business centers, insider threats, or the classic “evil maid” attack where a device is briefly left unattended but powered off or in sleep mode.

Impact Far Beyond Gaming

While the issue came to light via anti-cheat enforcement, its implications extend far beyond gaming. Any malware that can gain control this early can leverage the vulnerability to:

  • steal credentials and cryptographic keys stored in memory;
  • deploy persistent bootkits that survive OS reinstalls;
  • silently modify security drivers and monitoring agents.

This makes the vulnerability highly relevant for enterprise environments, especially where secure boot chains, disk encryption and strong endpoint integrity guarantees are central to the organization’s threat model.

Mitigation Steps for Users and Security Teams

Security professionals recommend that owners of systems based on Asus, Gigabyte, MSI and ASRock motherboards promptly check for UEFI/BIOS updates on their vendor’s support site. Before flashing firmware, users should back up important data and follow the manufacturer’s instructions carefully to avoid misconfiguration or system instability.

In corporate and high-value environments, it is advisable to:

  • establish centralized inventory and lifecycle management for firmware versions and apply security updates routinely;
  • enable and enforce IOMMU/VT-d, Secure Boot and TPM-based protections wherever supported;
  • restrict physical access to workstations and servers, especially open PCIe slots and Thunderbolt ports;
  • use full disk encryption with pre-boot authentication and integrity checks for the boot chain;
  • monitor vendor advisories and alerts from coordination centers such as CERT/CC for new UEFI and firmware-related vulnerabilities.

The disclosure of this UEFI DMA vulnerability underscores that application- and OS-level security is not sufficient if the underlying platform firmware is misconfigured. Regular firmware updates, rigorous control of physical and port access, and close attention to security warnings from tools like Vanguard are essential to reducing the risk of both cheating in online games and far more serious compromises of corporate infrastructure. Organizations and individual users alike should treat platform firmware as a critical security component and keep it under the same continuous scrutiny as operating systems and applications.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.