The vulnerability in Arm ArmNN through 2026-03-27 involves an integer overflow in the TensorShape::GetNumElements() function in armnn/Tensor.cpp. The overflow happens because tensor dimension multiplication uses 32-bit unsigned arithmetic without checking for overflow, causing the GetNumBytes() function to return a smaller-than-actual allocation size. Consequently, during the Optimize()->InferOutputShapes() process, the BatchToSpaceNdLayer reads beyond the allocated buffer, resulting in a heap-based buffer over-read. This can be triggered by a specially crafted TFLite model file that exploits the flawed size validation.

This vulnerability can lead to a heap-based buffer over-read, which may cause application crashes or potentially expose sensitive memory contents. There is no information about known exploits in the wild. The impact is limited to the affected versions of Arm ArmNN prior to 2026-03-27. No CVSS score is provided, and no vendor advisory or patch information is available to confirm remediation status.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, avoid processing untrusted or malformed TFLite model files with affected versions of Arm ArmNN. Monitor vendor channels for updates and apply patches once released.