Secure and Resource-Efficient Deployment of LLMs Using Confidential Computing

Problem statement

Even protected inputs are processed on infrastructure that may require trust in cloud/container operators, administrators and privileged components.

Objective

Build and benchmark an isolated LLM execution architecture using containers and confidential computing, focusing on security, latency and resource efficiency.

Student focus

  • Confidential computing
  • Benchmarking
  • Secure deployment
Tags: Cybersecurity, Cyberphysical, AI
Published Oct. 13, 2026 11:24 AM - Last modified Oct. 13, 2026 11:24 AM

Supervisor(s)

Scope (credits)

30