Research papers’ supplemental materials authenticity analyzed with Confidential Computing utilizing Intel TDX

Abstract

Confidential computing is an emerging technology that enables secure computation by ensuring data confidentiality, data integrity, and code integrity—unauthorized entities cannot view, alter, or manipulate data or code during processing. To assess the reliability of published research, our team used confidential computing, specifically Intel Trust Domain Extensions (TDX), to analyze the accuracy of supplemental materials in research papers from Management Science. The motivation was to ensure that published code remained unaltered and secure from external interference during replication, and to check for its accuracy in the paper. Using a virtual machine deployed on Microsoft Azure with free credits, we executed code in a confidential environment. This approach not only assured computational integrity but also demonstrated improved performance. Notably, the confidential computing environment ran approximately 10% faster than a non-secure setup, though this may reflect efficiencies from the virtual machine rather than TDX itself. Our findings show that confidential computing enhances the trustworthiness of research reproducibility efforts by shielding code from tampering and ensuring faithful execution. However, one limitation is that TDX only secures locally executed applications—web-based platforms like SAS OnDemand or free versions of MATLAB are excluded. We conclude that confidential computing is a promising tool for auditing and verifying research, offering both performance gains and security, strengthening reproducibility across scientific disciplines.