CWE-1190 describes a product that enables a Direct Memory Access (DMA) capable device before the security configuration is established. An attacker can then extract data from the product or gain privileges on it. It concerns hardware and firmware teams who define boot order.
How it happens
DMA lets a device move data using direct hardware access to main memory, without the operating system in the middle. That speed is the point, and also the risk. A DMA device can read or write memory without passing through OS access control.
Several virtualization based mitigations exist against DMA attacks, and they are usually configured at boot time. The trouble is that some IP blocks, known as early boot IPs, power up before boot is complete and may be DMA capable. If such an IP is not trusted, it could launch a DMA attack and reach assets that should otherwise be protected, because the protections are not in place yet.
What an attacker gains
The record lists a bypass of a protection mechanism with the ability to modify memory, rated high impact. Because DMA devices have direct write access to main memory, and because the attack happens early, the attacker can get around OS or bootloader access control.
How to find it
The record lists no detection methods. A useful review is to map the boot sequence and ask, for each DMA capable block, what protects memory at the moment that block becomes active.
How to fix it
MITRE's mitigation is to use an IOMMU to orchestrate IO access from the start of the boot process. The protection has to exist before the device can act, not after.
Quick check
Why is early boot the dangerous window?
Memory protections against DMA are usually configured during boot. A DMA capable block that powers up before that configuration can access memory without the protections in place.
