Academic researchers have disclosed DDRop, a hardware attack that defeats the memory-integrity guarantees of Intel and AMD confidential computing by silently dropping writes to a server’s DDR5 memory, so the processor keeps reading stale encrypted data as if it were current. The technique works against Intel TDX, Intel Scalable SGX, and AMD SEV-SNP, the technologies cloud providers use to keep customer workloads encrypted in memory even from the provider itself. Intel confirmed the research in its own security announcement. “Intel’s analysis confirms that the described scenarios fall outside Intel’s standard threat model for confidential computing deployments,” the company said, adding it is “evaluating additional architectural hardening options and detection mechanisms.”
The attack requires an adversary who already has privileged software access to the target server and brief physical access to install a hardware interposer between the processor and a memory module. It is not remotely exploitable, which matters for how seriously to weigh it: DDRop is not a network threat to cloud tenants today, but it is a real gap in a model many organizations treat as a hard guarantee against a malicious or compromised cloud provider or a data-center insider, a trust assumption CyberTech has flagged before in the context of memory-resident compromise techniques that hide from standard inspection.
The original insight is in what confidential computing’s design deliberately left out to make it practical at cloud scale. Intel and AMD’s memory encryption confirms that data is encrypted, but omits a “freshness” guarantee confirming the processor reads the most recently written value rather than a stale one, since verifying freshness across a full server’s memory is expensive. DDRop is the first active interposer attack to exploit that gap against current DDR5 hardware, rather than only passively reading it as the earlier TEE.fail technique did. Intel’s recommended mitigations, restricting physical server access and applying least-privilege controls on system software, are defense-in-depth practices organizations should already run, which is itself the point: confidential computing adds a layer on top of physical and supply-chain security, it does not replace it, a distinction that matters as more machine-identity and infrastructure-trust gaps surface, including CyberTech’s recent coverage of certificate-management failures in machine-identity infrastructure.
Source: Intel
