Confidential Computing and Trusted Execution Environments: A Survey of Technologies

Abstract

Conventional security controls protect data at rest through storage encryption and data in transit through transport encryption, yet data in use, meaning plaintext held in memory and registers during computation, has remained exposed to any sufficiently privileged component of the software stack. Confidential computing addresses this gap by performing sensitive computation inside a hardware-based, attested Trusted Execution Environment (TEE) that isolates code and data even from the operating system, hypervisor, and cloud operator. This survey reviews the motivation, threat model, and technology landscape of confidential computing. It examines process enclaves such as Intel SGX alongside confidential virtual machine designs including Intel TDX, AMD SEV and SEV-SNP, Arm TrustZone with the Confidential Compute Architecture, and the RISC-V Keystone framework, comparing them across isolation granularity, memory encryption, attestation, and vendor. It describes remote attestation and hardware roots of trust, surveys applications spanning confidential artificial intelligence, privacy-preserving analytics, blockchain, key management, and multi-party computation, and positions the paradigm as a complement to homomorphic encryption, secure multi-party computation, and differential privacy rather than a replacement. A dedicated security analysis reports side-channel and transient-execution attacks such as Foreshadow, Plundervolt, and SGAxe, together with representative performance overheads, and the review closes with standardization efforts and future directions including confidential graphics processing units.