CVE-2025-0819: Arm 5th Gen GPU Driver Use After Free Flaw

CVE-2025-0819 Overview

CVE-2025-0819 is a Use After Free vulnerability affecting multiple Arm GPU kernel drivers, including the Bifrost GPU Kernel Driver, Valhall GPU Kernel Driver, and Arm 5th Gen GPU Architecture Kernel Driver. This memory corruption flaw allows a local non-privileged user process to perform valid GPU memory processing operations to gain access to already freed memory, potentially leading to privilege escalation or arbitrary code execution.

Critical Impact

Local attackers with low privileges can exploit GPU memory operations to access freed memory, potentially achieving high impact on confidentiality, integrity, and availability of affected systems.

Affected Products

• Arm Bifrost GPU Kernel Driver versions r44p0 through r49p3 and r50p0 through r51p0
• Arm Valhall GPU Kernel Driver versions r44p0 through r49p3 and r50p0 through r54p0
• Arm 5th Gen GPU Architecture Kernel Driver versions r44p0 through r49p3 and r50p0 through r54p0

Discovery Timeline

• 2025-06-02 - CVE-2025-0819 published to NVD
• 2025-07-02 - Last updated in NVD database

Technical Details for CVE-2025-0819

Vulnerability Analysis

This Use After Free (CWE-416) vulnerability resides in the memory management subsystem of Arm's GPU kernel drivers. The flaw occurs when GPU memory operations continue to reference memory regions that have already been deallocated. When a local user process performs legitimate GPU memory processing operations, the driver fails to properly validate that the memory being accessed is still allocated and valid.

The vulnerability enables an attacker to manipulate the contents of freed memory through carefully crafted GPU operations. Since the memory has been freed but is still accessible through dangling references, an attacker can potentially overwrite critical data structures or function pointers that may be reallocated to the same memory region.

Root Cause

The root cause of CVE-2025-0819 lies in improper memory lifecycle management within the GPU kernel drivers. The driver does not adequately track or invalidate references to GPU memory buffers after they are freed. This creates a race condition where legitimate GPU operations can access stale memory references, allowing the exploitation of dangling pointers.

The affected drivers—Bifrost, Valhall, and 5th Gen GPU Architecture—share common memory management code paths that fail to synchronize memory deallocation with active GPU operations, leaving windows for exploitation.

Attack Vector

The attack is executed locally and requires the attacker to have low-privileged access to the system. No user interaction is required for exploitation. The attacker can craft a sequence of GPU memory operations that:

1. Allocates GPU memory buffers through the kernel driver
2. Initiates GPU operations that reference these buffers
3. Frees the memory buffers while operations are still pending or while references remain
4. Triggers memory reuse that allows controlled data to occupy the freed region
5. Leverages the stale references to read or write to attacker-controlled memory

The vulnerability allows local privilege escalation by corrupting kernel data structures, potentially granting the attacker elevated privileges on the affected device.

Detection Methods for CVE-2025-0819

Indicators of Compromise

• Unusual GPU memory allocation patterns with rapid allocation and deallocation cycles
• Kernel panic or crash logs referencing Arm GPU driver modules (mali_kbase or related kernel modules)
• Unexpected privilege escalation events from non-privileged user processes
• Memory corruption signatures in kernel logs associated with GPU subsystem operations

Detection Strategies

• Monitor kernel logs for GPU driver-related errors, particularly memory access violations or null pointer dereferences
• Implement system call monitoring for GPU ioctl operations that show abnormal memory management patterns
• Deploy endpoint detection solutions capable of identifying exploitation attempts targeting GPU kernel drivers
• Use kernel integrity monitoring to detect unauthorized modifications to kernel memory regions

Monitoring Recommendations

• Enable enhanced kernel auditing for GPU driver operations on devices using Arm Mali GPUs
• Configure alerts for processes exhibiting unusual GPU memory allocation behavior
• Monitor for attempts to load or interact with vulnerable driver versions (r44p0 through r54p0 depending on driver family)
• Implement SentinelOne Singularity platform for real-time kernel-level threat detection on affected devices

How to Mitigate CVE-2025-0819

Immediate Actions Required

• Update Arm GPU kernel drivers to patched versions beyond the affected ranges as specified by Arm
• Audit systems to identify devices running vulnerable driver versions
• Restrict local access to sensitive systems until patches can be applied
• Enable additional kernel hardening features such as KASLR and CFI where available

Patch Information

Arm has released updated driver versions that address this vulnerability. Affected organizations should consult the ARM Documentation Version 110466 for specific patch information and remediation guidance. The patch resolves the Use After Free condition by implementing proper memory reference tracking and validation during GPU memory operations.

For Bifrost GPU Kernel Driver, update to versions after r51p0. For Valhall and 5th Gen GPU Architecture Kernel Drivers, update to versions after r54p0.

Workarounds

• Limit local user access to systems with vulnerable GPU drivers until patches can be deployed
• Implement application sandboxing to restrict GPU access from untrusted applications
• Consider disabling GPU acceleration in high-security environments as a temporary measure
• Apply principle of least privilege to reduce the attack surface for local privilege escalation