Intel TDX Module CVE-2026-20705

Exploitation requires all of the following to be true simultaneously: (1) The target system must be an Intel platform with TDX support enabled in firmware and an active Trust Domain workload running. (2) The attacker must have achieved Ring-0 (kernel/hypervisor) privileged code execution on the host - equivalent to full hypervisor compromise or host OS root with hardware access. (3) The attack must meet specific preconditions indicated by AT:P in the CVSS 4.0 vector; the description references 'attack requirements are present' without disclosing the exact trigger condition - consult INTEL-SA-01436 for specifics. (4) High attack complexity (AC:H) applies: the attacker cannot trivially exploit this without a tailored, non-generic approach. Factors limiting exploitation include: physical host local access only (no network vector), requirement for pre-existing Ring-0 privilege escalation, hardware dependency on TDX-capable Intel platforms, and the need for specific (undisclosed) attack prerequisites. No user interaction by a TD guest is required once prerequisites are met.

The CVSS 4.0 score of 6.8 (Medium) reflects the compounding mitigating factors in the attack requirements: local access only (AV:L), high attack complexity (AC:H), specific attack prerequisites (AT:P), and high privilege level (PR:H). An adversary must already operate as privileged system software at Ring 0 on the target host - a position that typically implies an already-compromised hypervisor or host OS. The SC:H (subsequent system confidentiality: high) component is the most damaging signal: successful exploitation breaches the TDX isolation boundary, potentially exposing secrets from guest Trust Domains to the host. EPSS data was not provided; given the narrow exploitation conditions and specialized hardware dependency, widespread automated exploitation is unlikely. No KEV listing exists and no POC has been identified. The threat model is primarily relevant to cloud multi-tenant environments where a privileged host-side actor (e.g., a rogue cloud operator or an attacker who has fully compromised the hypervisor) targets TDX-protected tenant workloads. For organizations not deploying TDX-based confidential VMs, this vulnerability has no practical impact. For confidential computing deployments, the trust model assumption (untrusted hypervisor) makes this a meaningful integrity concern despite the high bar for exploitation.

A threat actor who has achieved privileged code execution at the hypervisor or host OS kernel level (Ring 0) on an Intel TDX-capable server crafts a specific, high-complexity interaction with the TDX module that triggers insecure memory storage behavior - potentially reading sensitive TD secrets, cryptographic keys, or attestation data from a running Trust Domain tenant workload. No public proof-of-concept code has been identified at time of analysis, and replicating the attack requires specialized knowledge of TDX internals combined with an existing high-privilege foothold on the physical host.