Research roundup documents years of attacks breaking TEE security guarantees
A new analysis compiles a series of publicly documented attacks against trusted execution environments (TEEs) from Intel SGX and AMD SEV, including Foreshadow, SGAxe, ÆPIC Leak, BadRAM and WireTap.fail. These attacks range from software bugs in enclave code, like a memory-safety flaw in Signal's contact discovery enclave, to hardware-level exploits that let attackers forge or steal remote attestation keys, sometimes using physical access to memory chips.
GoKawiil's interpretation of the reporting above, not reported fact.
The compilation suggests that TEE vendors have struggled to make durable security guarantees despite marketing them as hardware-rooted trust anchors, since side-channels and attestation forgeries keep resurfacing across multiple vendors and years. This could prompt developers and companies relying on TEEs for confidential computing, secure messaging, or cloud isolation to reassess their threat models or reduce dependence on hardware attestation as a sole safeguard. The author argues the complexity of modern CPUs makes such security claims inherently difficult to verify or trust fully.
- Multiple real-world attacks (Foreshadow, SGAxe, ÆPIC Leak, BadRAM, WireTap.fail) have broken attestation security in Intel SGX and AMD SEV over the past several years.
- Some attacks stem from bugs in enclave application code itself, such as a memory-safety flaw in Signal's contact discovery enclave.
- Other attacks require physical hardware access, such as reprogramming DIMM chips, raising questions about the real-world threat model for TEE deployments.
Source: alinush.github.io — Alin Tomescu, 2026-10-18
Published there as: “It's time to talk about TEEs”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.